GitLab

at 2012-11-23T09:35:49-08:00

First cut at ARM vm definitions.

First step at insts. Incomplete, but testing out pull requests.

Add instruction definitions for basic data-processing, load/store, and
branches.

Clean up so it can compile.

Add ARM package.

Update vmdir to add arm directory.

Get file to compile successfully.

o Add PC as a defreg, so we can do (simple) PC-relative addressing.
o Clean up locations for float types.

Merge branch 'rtoy-update-packages-setup' into add-insts-1

First version of parms for ARM, based on SPARC.

Add arm-fasl-file-implementation.

Merge branch 'rtoy-add-parms' into add-insts-1

Add default value for cond for the branch instructions.

Export arm-fasl-file-implementation.

Add multiply instructions.

Update from review.

More instructions
o Cleaned up some things, with commments added.
o Divide and BKPT instructions added.
o Added many floating-point instructions.

Compiles, with warnings.

o Clean up fp-reg-tn-encoding
o Other random cleanups and typo fixes.

o Add format-base for the basic instruction layout and try to use it
  in more places.
o Remove disassembler instruction conflicts from all instructions
  except ldrh and ldrsh. (Don't know why they conflict.  They look
  different.)

Add movw and movt.

Change format-mul and format-bkpt to include format-base.

Change format-2-immed and format-2-reg to include format-base.

o Change format-2-halfword* to format-0-halfword*, and include
  format-base.
o Fix some issues in conflicting instructions because we forgot to
  set the size field for the disassembler.
o Actually implement the vmov instructions.

Woohoo!  disassem::print-backend-inst-space prints out all the defined
instructions.

Update printer for condition-code argument type to print out the cond
field correctly.  In all other inst printers, we can just say "cond"
now.  (I think.)

Fix the encodings for single and double regs.

o Define VLDR/VSTR instructions.
o Minor cleanup for printing BLX instructions.

o Add shift-type arg type and use it to print out the shift type. 
o Add dummy maybe-add-notes.
o Correct some printers to print the right thing.

Fix some more printers.

o Fix up printers for many instructions
o Fix up some instruction formats
o Fix up some instrcution definition macros

o Clean up the MOV inst.  It doesn't support the shifted register
  operand anymore. (That's a pseudo-op according to ARM).
o Added the shift instructions.

Fix printing of shift instructions.

Update VCMP/VCMPE instruction so that it at least disassemble.

Many bug fixes:
o There is no STRSH and STRSB instructions, Duh!
o Make all instruction formats include format-base and update
  emitters and instructions appropriately.
o Define MVN separately from the other data processing instructions;
  it's not quite like them.
o Fix some bugs in bits in the shift instructions.
o Fix bad definition of mls (it never sets the flags)

Add UDF, SVC, and NOP instructions.

Add vmsr and vmrs to access FPSCR.

o Add MRS
o Add VMOV (ARM reg to/from single-reg)

Initial version of parms, copied from the sparc port.

Merged in arm-rtoy-add-parms (pull request #4: Initial version of parms, copied from the sparc port.)
Update parms for ARM

o Add correct float trap info for ARM.
o Add initial guess for static symbols, number-stack-displacement.
o Remove sparcisms.

Support fixups for the MOVW and MOVT instructions.  (Untested.)

Clean up
o Add some comments,
o Indent some things

Add MSR and fix some issues
o Add MSR instruction
o Fix incorrect number of args for emit-format-vfp-fpscr.
o Fix incorrect number of args for emitter in define-load/store-extra.

o Update inst column width
o Try to clean up VCVT instructions somewhat. Still needs work.

o Need to use STREAM arg in SHIFT-TYPE arg-type printer.
o Add CLZ instruction.

o Allow using ARM register names instead of lisp register names when
  disassembling instructions.  (This is useful when comparing
  disasembly from other disassemblers that use ARM syntax.)
o Minor changes to use less vertical code space.
o Forgot to add #<shift> for ROR shifts.
o Forget to set z 0 for the format-0-reg-shifted printer.

Change BKPT to allow any cond value in disassembly, but only allow :al
on assembly.

Add instruction macro li.

Update according to review.

Merged in rtoy-update-packages-setup (pull request #3: Add ARM package, update vmdir, get vm.lisp to compile)
Initial version of script to cross-compile from x86 to arm/linux.

Remove comment that should have been removed.  Pull request #3 didn't
seem to merge this correctly?

Turn off scheduler; update some dependency information.

Merge branch 'arm-rtoy-add-insts' into arm-rtoy-add-macros

Conflicts:
	src/compiler/arm/vm.lisp

Merge branch 'master' into arm-rtoy-add-insts

Conflicts:
	src/compiler/arm/vm.lisp

Merge branch 'arm-rtoy-add-insts' into arm-rtoy-add-macros

Initial checkin of sparc version.

Add macros.lisp.

Fix according to review comments.

Merged in arm-rtoy-add-parms (pull request #5: Update parms for ARM)
Merge branch 'master' into arm-rtoy-add-insts

Conflicts:
	src/compiler/arm/parms.lisp

Update to match current master version.

Clean up.

More minor clean ups.

Merge branch 'arm-rtoy-add-insts' into arm-rtoy-add-macros

Conflicts:
	src/compiler/arm/macros.lisp
	src/compiler/arm/parms.lisp

Default condition value in INST-CONDITION-CODE was wrong.

Merge branch 'arm-rtoy-add-insts' into arm-rtoy-add-macros

Fix typo, remove long-float-digits and exports for long-float stuff.

Add missing closing paren, get rid of long-float-digits and exported
long-float symbols.

Merged in arm-rtoy-fix-parms (pull request #7: Fix typo and remove more long-float stuff)
Merge branch 'master' into arm-rtoy-add-macros

Fix some comments; add FIXME notes.

Merge branch 'master' into arm-rtoy-add-insts

o Fix according to review.
o EMIT-FORMAT-VFP-2-ARG had too few entries. Update all uses of this
  emitter too.
o FORMAT-VFP-2-ARG was missing (byte 3 25).
o VMOV of ARM core registers was missing the cond field in the
  emitter.
o FORMAT-VFP-FPSCR was missing (byte 3 25).  Update users
o VMRS was missing an arg in the emitter.

Fix SC stuff VMOV (for moving between ARM and FP regs).

Fix according to review.

Update from review comments. More to come.

Forgot the cond field for VMRS and VMSR.

More cleanup from review.

More updates from review:
o Rename DEFEINE-DATA-PROC
o Split data processing instructions to include two separate
  instructions:  the basic inst and the one that sets the flags.
o Define new function to convert the condition flag to the appropriate
  code, with error checking.

Move body of emitter of define-one-data-proc-inst to its own function.

o Use EMIT-DATA-PROC-FORMAT in more places
o Update it to handle the case where src1 is 0.

o Need to define src1 as a reg type to disassemble compare
  instructions correctly.
o Update EMIT-DATA-PROC-FORMAT to handle the case where dst should be
  0 instead of a tn.

Set opcode width to 8; 16 is too long for most instructions.

Add closing paren that was missing.

More cleanups to have separate instruction definitions for one with no
setting of flags and one with.

Fix disasembly for umull and other two-output multipliers.  The
registers were printed in the wrong order.

o Remove another use of inst-condition-code
o NOP instruction was not printing out the cond field.

Set opcode column width to 9 for umullsne, and similar.

Test code for disassembler.

Add comments.

Add docstrings for make-shift and make-ea.

Fix typos introduced when changing :post-indexed to :indexed.

More cleanups.
o Replace INST-CONDITION-CODE with CONDITION-CODE-ENCODING and remove
  that function.
o Remove INST-SET-FLAGS and fix up one last usage of it.
o Rename ENCODE-SHIFT to SHIFT-TYPE-ENCODING to match
  CONDITION-CODE-ENCODING.

Update according to review comments.

Update according to review

insts.lisp:
o Update according to review.
o Fix some bugs in some instructions

test-disassem.lisp:
o Update some disassembly tests to include the fact that we now have
  separate <foo> and <foo>s instructions.

Merged in arm-rtoy-add-insts (pull request #6)

Add ARM instructions, cross-compile script, and a simple test program for the disassembler.
Fix up FP insts that were postponed.

arm/insts.lisp:
o Fix up FP insts
o Correct the printer for the VMOV.F32 and VMOV.F64 instructions.

arm/test-disassem.lisp:
o Add tests for the VMOV.F32 and VMOV.F64 instructions.

* Fix type issue in MAKE-SHIFT: SHIFT TYPE can also be :rrx.
* Change DEFINE-VFP-3-INST not to need inst-name.  Derive the type
  from the types of the args.

Simplify instruction definitions for vneg and vcmp and friends.  These
instructions now figure out the appropriate bits from the types of the
registers.

Also change vcvt instructions to have their own macro.

Oops.  Accidentally duplicated define-vfp-3-insts.  Remove it.

Remove unused stuff:

* Remove opa1 and &optional doublep from define-vfp-3-inst.
* Remove opc4 from define-vfp-2-inst.
* Fix typo in default printer and print name in define-vcvt-inst
* Remove opc3 from define-vfp-cmp-inst.

Add test pattern for vadd.

Fix typo.

Fix according to review comments.

* Fix indentation and whitespace issues.
* Fix mistake in load/store printer where the address was printed as
  [rn#+off] instead of [rn, #+off]
* Change define-fp-load/store-inst to derive the instruction size from
  the type of dst tn.

Add some tests for VLDR and VSTR.

Cleanup some indentation and use backquotes everywhere as needed.

Unify the VMOV variants into one instruction.

Add test for moving 2 ARM regs to/from a double-precision reg.

Cleanup according to review comments.

Also removed the redundant op0 field from
format-6-vfp-vmov-core-double.  It should have been derived from
format-base.  The printers were updated to use opb0 instead op0.

Merged in arm-rtoy-fixup-fp-insts (pull request #8)

Fix up FP insts that were postponed.
Merge branch 'master' into arm-rtoy-add-macros

Update according to Robert's review comments.

Replace 3 with (1- vm:word-bytes)

Initial version from sparc.

Merge branch 'arm-rtoy-add-macros' into arm-rtoy-add-support

Dummy stub implementation generating an error for everything.

Merge tag 'snapshot-2013-07' of ssh://common-lisp.net/var/git/projects/cmucl/cmucl into arm-rtoy-sync-2013-07

Snapshot 2013-07

Merged in arm-rtoy-sync-2013-07 (pull request #10)

Sync ARM code with upstream cmucl snapshot 2013-07
Add FIXME for Load-Stack-TN.

Update according to review comments.

Merged in arm-rtoy-add-macros (pull request #9)

Add macros.lisp, based on the sparc version
Merge branch 'master' into arm-rtoy-add-support

Fix package.

Initial version, copied from sparc.

Oops.  ARM's move instruction is called "mov", not "move".

Add some instructions for dumping data and header objects that we
didn't implement earlier.

Implement for ARM, but many VOPs signal errors.  Only the most obvious
VOPs are implemented.

o Fix typo move->mov
o Add not-implemented pseudo-instruction to cause a halt trap.

Use not-implemented.

Merge branch 'arm-rtoy-add-support' into arm-rtoy-add-move

o Use NOT-IMPLEMENTED inst instead of ERROR.
o Remove SPARC stuff

Remove bodies of vops which aren't implemented.

Fix typo: missing closing paren.

Fix another typo in :type arg of a structure slot.

Initial version, based on sparc/float.lisp

Fix some bugs
o Reindent def-load/store-word nicely
o In ALLOCATION, make sure instructions have three args; the ARM port
  doesn't support two-arg instructions so both source args are needed.
o In WITH-FIXED-ALLOCATION, we need to give PSEUDO-ATOMIC a temp reg.
o Assert that temp-tn is always given in PSEUDO-ATOMIC because we
  always need a temp reg.

Fix some bugs.
o The compare instructions don't have a dst arg, so remove it.
o Fix typo in the declaration of the type of src.  We wrote dst, not
  src as it should be.

First cut at an ARM implementation.  Not yet finished with all the
VOPS yet.

Disable :complex-fp-vops on ARM for now.

Implemeent more VOPS or make them not-implemented.  Long-float support
ripped out.

Remove misplaced extra closing paren.

Merged in arm-rtoy-add-support (pull request #11)

Add assembly/support.lisp, based on sparc
Merged in arm-rtoy-add-move (pull request #12)

Add arm/move.lisp, based on sparc
Merged in arm-rtoy-add-support (pull request #15)

Merge pull request #12 to master
Patch from the arm-rtoy-add-sap branch used in pull request #14, but
only includes sap.lisp, since that branch was based off of the float
branch.

Merged in arm-rtoy-add-sap-2 (pull request #16)

Implement sap.lisp (again)
Update according to review comments.

Change how ldr/str/vldr/vstr instructions are handled using new
approach.  MAKE-EA is gone.

* insts.lisp
  * Replace old MAKE-EA function and structures with PRE-INDEX,
    POST-INDEX, and MAKE-OP2.
  * Fix a couple of errors in vldr/vstr:
    * Addressing options are more limited than ldr
    * The offset is now explicit, and is checked for valid values.

* macros.lisp:
  * Fix some errors in the instructions because all 3 args are needed
    for three-arg instructions.
  * Update uses of MAKE-EA.
  * The or instruction is ORR, not OR.

* sap.lisp:
  * Update uses of MAKE-EA.

More fixups:
o Fix up definitions for compare instructions.
o Make load/store-extra-inst behave the same as load/store
  instructions.
o Fix typo in load/store.

More fixups:
o It's emit-format-0-reg-shifted, not emit-format-0-reg-shift
o It's src1, not imm16.
o It's emit-format-0-mul, not emit-format-mul.
o In load/store-extra-inst, make the index 8 bits, not 12.  We don't
  currently support the imm12 type.
o For load/store-inst, clean up implementation a bit
o Fix typo in udf inst:  segment, not eegment.
o Fix typo: emit-formt-0-msr -> emit-format-0-msr.
o Fix typo: format-vfp-3-arg -> emit-format-vfp-3-arg
o Fix quoting in define-vcvt-inst macro.
o It's dst-or-src, not dst-or-src-or-cond.

Fix more typos:
o indexing-mode-post-indexed -> indexing-mode-post-index
o encooding -> encoding

More fixes:
o Fix typo: reg -> tn
o Fix typo: opts -> cond
o Fix typo: emit-branch -> emit-branch-imm
o Add emit-header-data

Fix Issue #7.  Handle moving an immediate float value to a register.

Fix typo: emit-format-vfp-2-reg -> emit-format-vfp-2-arg.

Fix some typos:
o The type is called tn, not reg.
o The shift type is called shift-type, not shift-types.

Fix more typos:
o set-flags-p -> set-flags-bit.
o reg -> tn.
o There's no shift operand in the RRX instruction!

Fix a bunch more cases of reg -> tn.

No more warnings or issues when compiling insts.lisp. Yay!

ALLOCATION requires that a temp-tn be given because PSEUDO-ATOMIC
needs a temp-tn. Check for that in ALLOCATION and pass the temp-tn
through to PSEUDO-ATOMIC.  Check that a temp-tn is supplied in
PSEUDO-ATOMIC.

Add comments.

Import sparc/system.lisp.

Update for ARM, using NOT-IMPLEMENTED.

Import sparc char.lisp.

Update for ARM, using NOT-IMPLEMENTED.

Import sparc memory.lisp.

Update for ARM, using NOT-IMPLEMENTED.

Import sparc static-fn.lisp.

Update for ARM, using NOT-IMPLEMENTED.

Import sparc arith.lisp.

Fix boilerplate.

inst.lisp:
o Add NOT instruction macro

arith.lisp:
o First cut using NOT-IMPLEMENTED almost everywhere
o Remove the zero SC class.

o Forgot a few more generators that need to be NOT-IMPLEMENTED.
o Update logandc1/2, logior, and logxor with correct instructions.
  Comment out logorc1/2 and logeqv.

o Remove the binops that can't be done in one instruction.
o Remove sparc-v9 vops and references.
o Compile modular arith stuff in the ARM package, not C.

Remove modular versions of logeqv, logorc1, and logorc2 since we don't
have vops for those now.

Remove modular-backend versions of logeqv, logorc1, and logorc2 since
we don't have vops for those now.

Import sparc subprim.lisp.

Update for ARM, using NOT-IMPLEMENTED.

Import sparc debug.lisp.

Update for ARM, using NOT-IMPLEMENTED.

Import sparc c-call.lisp.

Fix boilerplate.

Remove long-float.

Update for ARM, using NOT-IMPLEMENTED.

Import sparc cell.lisp.

Update for ARM, using NOT-IMPLEMENTED.

Import sparc values.lisp.

Fix boilerplate.

Author: Raymond Toy <toy.raymond@gmail.com>
--- log message follows this line ---
Update for ARM, using NOT-IMPLEMENTED.

Import sparc alloc.lisp.

Fix boilerplate.

Author: Raymond Toy <toy.raymond@gmail.com>
--- log message follows this line ---
Update for ARM, using NOT-IMPLEMENTED.

Import sparc call.lisp.

Fix boilerplate.

Update for ARM, using NOT-IMPLEMENTED.

Import sparc nlx.lisp.

Fix boilerplate.

Update for ARM, using NOT-IMPLEMENTED.

o Fix up according to review comments.
o Add instruction macro for neg.

Clean up according to review comments.

Merge upstream CMUCL snapshot-2013-09 into arm-rtoy-sync-2013-09

Merged in arm-rtoy-sync-2013-09 (pull request #19)

Sync ARM code with upstream cmucl snapshot 2013-09
Merge branch 'master' into arm-rtoy-change-make-ea-2

Merged in arm-rtoy-change-make-ea-2 (pull request #18)

Remove MAKE-EA and replace it with a new scheme, take 2
Merge branch 'arm-rtoy-add-float' into arm-rtoy-add-float-2

First cut; float.lisp unchanged so it has errors.

o Clean up using the new vldr/vstr instruction.
o Anything vldr/vstr that didn't have a word offset has been replaced
  with NOT-IMPLEMENTED.

Update according to review.

Merge /Volumes/share2/src/clnet/cmucl/release-20e into arm-rtoy-sync-to-20e

Merged in arm-rtoy-sync-to-20e (pull request #21)

Sync ARM code with upstream cmucl 20e
Merge branch 'master' into arm-rtoy-add-float-2

Oops.  Replace move-long-reg with move-double-reg in a couple of
places.

Merged in arm-rtoy-add-float-2 (pull request #20)

Add arm/float.lisp, take 2
Implement system.lisp

Merge branch 'arm-rtoy-change-make-ea-add-char' into arm-rtoy-add-char.lisp

Merge branch 'arm-rtoy-change-make-ea-add-memory' into arm-rtoy-add-memory.lisp

Merge branch 'arm-rtoy-change-make-ea-add-arith' into arm-rtoy-add-arith.lisp

Merge branch 'arm-rtoy-change-make-ea-add-subprim' into arm-rtoy-add-subprim.lisp

Merge branch 'arm-rtoy-change-make-ea-addstatic-fn' into arm-rtoy-add-static-fn.lisp

Merge branch 'arm-rtoy-change-make-ea-debug' into arm-rtoy-add-debug.lisp

Merge branch 'arm-rtoy-change-make-ea-c-call' into arm-rtoy-add-c-call.lisp

Remove nl1-offset.

Merge branch 'arm-rtoy-add-memory.lisp' into arm-rtoy-add-char-mem-arit-subprim-system-static-fn-debug-c-call

Merge branch 'arm-rtoy-add-arith.lisp' into arm-rtoy-add-char-mem-arit-subprim-system-static-fn-debug-c-call

Merge branch 'arm-rtoy-add-subprim.lisp' into arm-rtoy-add-char-mem-arit-subprim-system-static-fn-debug-c-call

Merge branch 'arm-rtoy-add-system.lisp' into arm-rtoy-add-char-mem-arit-subprim-system-static-fn-debug-c-call

Merge branch 'arm-rtoy-add-static-fn.lisp' into arm-rtoy-add-char-mem-arit-subprim-system-static-fn-debug-c-call

Merge branch 'arm-rtoy-add-debug.lisp' into arm-rtoy-add-char-mem-arit-subprim-system-static-fn-debug-c-call

Merge branch 'arm-rtoy-add-c-call.lisp' into arm-rtoy-add-char-mem-arit-subprim-system-static-fn-debug-c-call

Merge branch 'arm-rtoy-change-make-ea-add-alloc' into arm-rtoy-add-more

Merge branch 'arm-rtoy-change-make-ea-add-call' into arm-rtoy-add-more

Merge branch 'arm-rtoy-change-make-ea-add-cell' into arm-rtoy-add-more

Merge branch 'arm-rtoy-change-make-ea-add-nlx' into arm-rtoy-add-more

Merge branch 'arm-rtoy-change-make-ea-add-values' into arm-rtoy-add-more

C callbacks aren't supported.

Import from sparc: emit-compute-inst, compute-code-from-lra.

Make emit-compute-inst and compute-code-from-lra using
NOT-IMPLEMENTED.

o Replace cname with lexenv.
o Remove a3-a5 offset.
o Remove nl1-offset.
o Remove nl1-nl3 offset.

Merged in arm-rtoy-add-more (pull request #31)

Add alloc.lisp, call.lisp, cell.lisp, nlx.lisp, values.lisp. 
Merge branch 'master' into arm-rtoy-add-char-mem-arit-subprim-system-static-fn-debug-c-call

Merged in arm-rtoy-add-char-mem-arit-subprim-system-static-fn-debug-c-call (pull request #30)

Pull requests 22-29 all in one
Copy sparc version.

Convert to ARM, using NOT-IMPLEMENTED.

Copy sparc version.

Convert to ARM, using NOT-IMPLEMENTED.

Copy sparc version.

Convert to ARM, using NOT-IMPLEMENTED.

Copy sparc version.

Convert to ARM, using NOT-IMPLEMENTED.

ARM doesn't have a zero SC.

Copy sparc version.

Convert to ARM, using NOT-IMPLEMENTED.

Copy sparc version.

Convert to ARM, using NOT-IMPLEMENTED.

Copy sparc version.

Convert to ARM, using NOT-IMPLEMENTED.

Copy sparc version.

Convert to ARM, using NOT-IMPLEMENTED.

Rip out non-existent non-descriptor regs.

o Rip out non-existent non-descriptor regs.
o Comment out some assembly routines because I don't know how to
  implement them even with NOT-IMPLEMENTED. See comments.

Merged in arm-rtoy-add-rest-assembly (pull request #33)

Add rest of the files in src/assembly/arm
Update according to review:

o Remove #-unicode and #+long-float.

Merged in arm-rtoy-add-rest-compiler (pull request #32)

Add rest of the files in src/compiler
Update according to review:

o Remove #-unicode and #+long-float.

Merged in arm-rtoy-add-rest-compiler (pull request #34)

Add rest of the files in src/compiler
Comment out fast-unary-ftruncate and double-float-bits. These are not
yet supported in the rest of the code for ARM.

Comment out symbol-hash and %set-symbol-hash. These are not yet
supported in the rest of the code for ARM.

Fix cross-compile issues. We can now build and load the
cross-compiler.

assembly/arm/array.lisp:
o Oops.  There is no a3 register. Randomly rename to a2 until we
  implement this.

compiler/arm/call.lisp:
o ARM doesn't support the clear-memory option when allocating a frame,
  so comment that out from the vop info args.

tools/cross-scripts/cross-x86-arm.lisp:
o Need to rename the cross-compiler package (X86) to OLD-X86.  Forgot
  to do that initially.
o Export the OLD-X86 names (instead of X86, which no longer exists
  since it was renamed).

tools/worldcom.lisp:
o Oops. Need to add case for ARM to compile the files from the
  assembly/arm directory.

Merged in arm-rtoy-cross-compile-fix (pull request #35)

Fixes to get cross-compiler to build and load.
Comment out the extra descriptor temps in DEFINE-FULL-CALL. This
causes a packing error because too many descriptors are used.

ldr/str instruction was not handling the case with an immediate offset
value. Add support.

It's lisp::*allocation-pointer*, not arm::*allocation-pointer*.

Just use a normal tn for nfp-tn.  (Not sure if this is right or not,
but we don't have a dedicated nfp register on ARM.)

Assembly works, if loaded up late enough.

Allow interactive mode where output goes to the screen and errors stop
the compilation.

Merge branch 'master' of /Volumes/share2/src/clnet/cmucl/dev/cmucl into rtoy-merge-2013-11

Merged in arm-rtoy-merge-upstream-2013-11 (pull request #37)

Merge cmucl 2013-11
More cross-compiler fixes

o Move macro NOT-IMPLEMENTED from macros.lisp to insts.lisp because
  macros.lisp is loaded AFTER insts.lisp.  Need the macro to be
  defined because we use NOT-IMPLEMENTED in a few places.
o Can't use NOT-IMPLEMENTED in COMPUTE-CODE-FROM-LRA because we need
  an emitter function, not an assembly instruction.  Blindly use the
  sparc version.

Fix typo.

Update according to review

o New make-nfp-tn
o Remove commented-out code.

ignore-me-sc-number doesn't exist.  Use something else for now.

sap-ref-single and sap-ref-double has to be implemented in a different
way because vldr is different from ldr.

Fix up according to review.

insts.lisp:
o Add type ARM-SIGNED-OFFSET.
o Use it.
o Make indentation consistent in a few places.

sap.lisp:
o Use ARM-SIGNED-OFFSET

Merged in arm-rtoy-xcompile-1 (pull request #36)

Cross-compile step 1
Add comment of where we can stop to play with the target compiler.

Add constant NUMBER-STACK-DISPLACEMENT, for consistency with sparc.
Correct the value or remove it later if not needed.

Implement COMPUTE-CODE-FROM-FN, based on sparc version.

Implement IF-EQ vop, based on sparc version.

Implement more of the guts of XEP-ALLOCATE-FRAME.

Fix many issues and implement a few more missing pieces.

The fixes:
o Forgot to define a printer for the reg MOV case.
o Use the correct values for the emitter for the register MOV inst.
o Comment out the flex-operand case. We don't support that.
o The split-imm16-printer was not printing the imm16 value out
  correctly. 
o Add some comments to the format-2-immed-printer to make it clearer
  what's happening.  (We had the U and W values backwards.)
o Rename DECODE-INDEXING-MODE to DECODE-IMMEDIATE-INDEXING-MODE to
  make it clearer.  Add some comments too.
o Add comments to DECODE-INDEXING-AND-OPTIONS.
o Fix DEFINE-LOAD/STORE-INST.
  o We were not generating the correct instruction for stores.
  o Need to compute the reg encoding for the indexing-mode register.
o Better implementation of NOT-IMPLEMENTED that takes an arg saves
  that to r0 so the trap handler can figure out what VOP is not
  implemented.

The new stuff:
o Implement EMIT-COMPUTE-INST better, and always use the worst-case
  for now.

Add defregtn for a0 so NOT-IMPLEMENTED can access it easily.

In the generator, define a macro VOP-NOT-IMPLEMENTED that calls
NOT-IMPLEMENTED with the vop name (as a symbol).

Don't need to load insts.lisp anymore since we're farther along.  Just
comment it out.

Implement more.

o In XEP-ALLOCATE-FRAME use the new VOP-NOT-IMPLEMENTED.
o Implement VERIFY-ARGUMENT-COUNT.

Add macrolet VOP-NOT-IMPLEMENTED to the generator so that VOPs can
call VOP-NOT-IMPLEMENTED to get the VOP name. VOP-NOT-IMPLEMENTED
calls NOT-IMPLEMENTED with the VOP name.

Replace NOT-IMPLEMENTED with VOP-NOT-IMPLEMENTED to pass the VOP name
to NOT-IMPLEMENTED.

Rename VOP-NOT-IMPLEMENTED to EMIT-NOT-IMPLEMENTED.  But still calls
NOT-IMPLEMENTED. 

Update all uses of VOP-NOT-IMPLEMENTED accordingly.

Merged in arm-rtoy-compile-1 (pull request #38)

Compile step 1
Fix typo.  The saves should not be in the assembl block.
(Found this when I tried compiling this on x86.)

Merged in rtoy-fix-emit-not-implemented (pull request #39)

Fix typo.  The saves should not be in the assembly block.
Merge tag 'snapshot-2014-03' of ssh://common-lisp.net/var/git/projects/cmucl/cmucl into rtoy-sync-snapshot-2014-03

Snapshot 2014-03

Merged in rtoy-sync-snapshot-2014-03 (pull request #40)

Sync to upstream snapshot-2014-03
Implement EMIT-NOT-IMPLEMENTED for assembly-routines and for
vm-support-routines.

There's no cname; use lexenv instead.

Missed this before; it's EMIT-NOT-IMPLEMENTED, not
VOP-NOT-IMPLEMENTED.

Merged in arm-rtoy-more-emit-not-implemented (pull request #41)

More EMIT-NOT-IMPLEMENTED fixes and a few typos
Use register layout assignment 3.

compiler/arm/vm.lisp:
o Use the register assignment 3 from the design notes doc.

compiler/arm/c-call.lisp:
o Use csp-offset instead of nsp-offset, which doesn't exist anymore.
o Remove reference to cfunc, which doesn't exist anymore.

compiler/arm/insts.lisp:
o Use csp-tn instead of nsp-tn for emit-not-implemented.

compiler/arm/parms.lisp:
o Add *c-stack-pointer* to hold the real C stack pointer since we're
  using the ARM SP (r13) register as the Lisp stack pointer.
o Remove the static FP constants.  ARM can load these as immediates.

compiler/arm/print.lisp:
o Remove reference to cfunc, which doesn't exist anymore.

Update according to review.

Use any tn for the number stack tn.

Remove the code that references nsp-tn.  We'll have to reimplement it
ourselves later since we don't have an nsp-tn anymore.

Merged in arm-rtoy-reg-assignments-3 (pull request #42)

Use register layout assignment 3.
Merge ssh://common-lisp.net/var/git/projects/cmucl/cmucl into arm-rtoy-sync-2014-09
to sync to latest upstream code.

Merged in arm-rtoy-sync-2014-09 (pull request #43)

Sync to upstream 2014-09 sources
Merge ssh://common-lisp.net/var/git/projects/cmucl/cmucl into rtoy-sync-2014-10-12

Merged in arm-rtoy-sync-2014-10-12 (pull request #44)

Sync to upstream sources 2014-10-12
Use the LIP register as the NFP register too.

Enable *debug-pack* and call lisp-conflicts when we fail to pack.

Comment out temporaries in vops where the temps area used.

If the temp was targeted, remove the targeting as well.

Comment out all temporaries from VOPs when possible.

Many of these, along with the args and results of the VOP were causing
TN conflicts during compilation.  It's silly to have them when most
VOPs aren't even implemented, so kill them off when possible.  Also
required moving some targeting of args/results to temporaries.

Remove the temps instead of commenting the out.

Oops. Really remove the temps instead of conditionalize them out.

Add more comments.

Merge branch 'arm-rtoy-nfp-using-lip' into arm-rtoy-remove-vop-temps

Conflicts:
	src/compiler/arm/call.lisp

Fix typo.  There is no x arg/result/temp in the MOVE-FROM-SIGNED VOP.

Undo *debug-pack* changes.

Address review comments and delete the #+nil'ed out code.

Replace the NFP register with the static symbol LISP::*NUMBER-FRAME-POINTER*.

Update all uses of NFP to use *number-frame-pointer*; use LIP as a
temporary, when needed, to load the value of *n-f-p*.

Add definition of lip-tn.

Merge branch 'arm-rtoy-nfp-using-lip' into arm-rtoy-remove-vop-temps

Oops.  *number-frame-pointer* isn't in the lisp package.  Remove
package qualifier.

Merge branch 'arm-rtoy-nfp-using-lip' into arm-rtoy-remove-vop-temps

Explictly declare a temp for lip instead of just using it.

Update comment.

Merge branch 'arm-rtoy-nfp-using-lip' into arm-rtoy-remove-vop-temps

Update comment about TN usage for MAKE-NFP-TN.

Merge branch 'arm-rtoy-nfp-using-lip' into arm-rtoy-remove-vop-temps

Fix typo in vm-support-routine's.

The VM support routines must return a list of the instructions.

Merged in arm-rtoy-fix-typo-generate-call-sequence (pull request #47)

Fix typo in vm-support-routine's
Merged in arm-rtoy-nfp-using-lip (pull request #45)

Add *number-frame-pointer* to be used as the nfp
Merge branch 'master' into arm-rtoy-remove-vop-temps

Merged in arm-rtoy-remove-vop-temps (pull request #46)

Remove temporaries from VOPs
Remove extraneous blank lines and #+nil stuff.

Merged in arm-rtoy-remove-vop-temps (pull request #48)

Remove extraneous blank lines and #+nil stuff.
Fix issue #17.

The load/store inst macro forgot to support the case where src2 can be
a simple register.

Fix issue #19: simplify disassembly for str/ldr.

If the shift type is LSL and the amount is 0, don't print anything,
since that's the default.

Set larger values for print length/level.

Use large values for *print-length*, *print-level*, and friends so
that compiler messages show more information.

Add lexenv to list of descriptor regs.

Forgot that lexenv is a descriptor reg and should be listed as
such. This Fixes issue #18.

Merge branch 'arm-rtoy-deeper-print-level-length' into arm-rtoy-issue-16

Merged in arm-rtoy-issue-17 (pull request #49)

Fix issue #17.
Merged in arm-rtoy-issue-18 (pull request #52)

Add lexenv to list of descriptor regs.
Move the print variables to the cross-compile lisp file instead of
cross-build-world.sh.

Merge branch 'master' into arm-rtoy-issue-19

Use macrolet.

Address comment 1.

Merged in arm-rtoy-deeper-print-level-length (pull request #51)

Set larger values for print length/level.
Add better comments as requested in review.

Remove temps from assembly routines.

Fixes issue #16.  There was a conflict between
allocate-vector and data-vector-ref/simple-vector.  Removed the other
temps from the assembly routines too, since they're not used yet.

Revert some changes accidentally pulled from other requests; they're
not relevant to this, but were useful for debugging this issue.

Merge tag 'snapshot-2014-12' of /Volumes/share2/src/clnet/cmucl/dev/cmucl into arm-rtoy-sync-2014-12

Snapshot 2014-12

Merged in arm-rtoy-sync-2014-12 (pull request #54)

Sync to upstream snapshot-2014-12
Merge branch 'master' into arm-rtoy-issue-16

Put back the code from pull request #49 for issue 17.

Fix issue #21: Handle ldrh rt, rn, 0

We weren't correctly handling the case where the third arg was an
immediate offset.

Add a test case for issue #21.

Add a function to disassemble the assembled result too so we can see
how it disassembles.

Handle negative immediate values correctly for ldrh.

Add simple test for this too.

Support EMIT-NOT-IMPLEMENTED in DEF-VM-SUPPORT-ROUTINE

This fixes issue #22 by wrapping the result returned from
GENERATE-RETURN-SEQUENCE in a macrolet defining EMIT-NOT-IMPLMENTED.

Merged in arm-rtoy-issue-19 (pull request #50)

Fix issue #19: simplify disassembly for str/ldr.
Merged in arm-rtoy-issue-16 (pull request #53)

Remove temps from assembly routines
Merge branch 'master' into arm-rtoy-issue-21

Address review comments.

Fix typos, clarify comments.

Replace EMIT-NOT-IMPLEMENTED in RECEIVE-UNKNOWN-VALUES

Fixes issue #23.  Just call NOT-IMPLEMENTED directly instead with the
appropriate symbol.

Replace EMIT-NOT-IMPLEMENTED in DEFAULT-UNKNOWN-VALUES and EMIT-MULTIPLY.

Cross compiler now builds without warnings.

Clean out armf files too.

See issue #24.

Merged in arm-rtoy-issue-24 (pull request #58)

Clean out armf files too.

Fixes issue #24.
Merge branch 'master' into arm-rtoy-issue-23

More fixups for ldrh.

The generated instruction is wrong when the second src operand is an
indexing operation. And there needs to be separate cases for src2
being an indexing operation and a simple TN. (For a simple TN, we were
subtracting instead of adding.)

Add tests for this as well.

Fix typo and line-up comments neatly for the disassembled instruction.

Arrange instruction examples in columns; clarify a comment.

Remove package prefixes from not-implemented.

Merged in arm-rtoy-issue-21 (pull request #55)

Fix issue #21: Handle assembly of ldrh with immediate or register offset
Merged in arm-rtoy-issue-22 (pull request #56)

Support EMIT-NOT-IMPLEMENTED in DEF-VM-SUPPORT-ROUTINE
Merged in arm-rtoy-issue-23 (pull request #57)

EMIT-NOT-IMPLEMENTED not defined in RECEIVE-UNKNOWN-VALUES, DEFAULT-UNKNOWN-VALUES and EMIT-MULTIPLY
Sort the list of fasl extensions alphabetically.

Merged in arm-rtoy-pull-request-58-a (pull request #59)

Sort the list of fasl extensions alphabetically.
Allow the prev and carry-in args to be on the unsigned-stack.

Merge branch 'master' into arm-rtoy-issue-25

Add OLD-VM:CHAR-BYTES.

See issue #27.

Merged in arm-rtoy-issue-25 (pull request #60)

Fix issue #25: Allow 2 args to be on the unsigned-stack
Merged in arm-rtoy-issue-27 (pull request #61)

Cross-compile script needs to add CHAR-BYTES
Put all int args on the stack instead in registers for call-out.

This follows what the RT port does in the call-out vop.

Also, some functions want two return values in registers, so update
RESULT-REG-OFFSET appropriately.

With these changes the cross-compile successfully builds a
kernel.core.

See issue #26.

Remove register-args slot from arg-state and simplify int-arg by
removing the unused branch that handled the integer register args.

Rename my-make-wired-tn to the more conspicuous c-call-wired-tn, like
for the RT port.

Merged in arm-rtoy-issue-26 (pull request #62)

Put all int args on the stack instead in registers for call-out.

Fixes issue #26.
Support ARM/linux config for create-target.sh

See issue #30.

Add Config files for ARM/linux.

See issue #29

When not using gencgc, we were selecting the wrong gc file and
options.  Follow sparc in this.

Add arm-lispregs.h and arm-validate.h

Initial implementation of these files.  Roughly correct but probably
needs work.  This is just to get things compiling.

Update according to comments.

Move -marm from CPPFLAGS to CFLAGS.
No need to check for gcc version.
Remove old __NO_CTYPE and _GNU_SOURCE defines.

Replace -mfpu=vfpv3-d16 with -ffp-contract=off to make sure we never
generate fused macs. This could potentially break the fdlibm routines.
vfpv3-d16 is the default anyway.

Update according to review.

Remove _ prefix (and suffix) from _ARM_LISPREGS_H_.
Define REGS macro which was left out.

Update according to review, and correct some sizes

Remove prefix and suffix from  _ARM_VALIDATE_H_ 
Use correct values for address map (from design doc).
Replace MB_128 with MB and adjust other usages appropriately.
Remove support for gencgc

Fix typo: sc -> scp.

Fix another typo: offset -> reg.

Fix up printf warnings.

See issue #34.

Address review comments.

Clean up CFLAGS some more; don't need -march or -fp-contract.

Set default heap size to 128 as specified in the design notes.  Also
detab and fix a few space issues.

Fix typo.

Add assert for nwords > 0 and add default: case.

Still need to initialize nwords to get rid of compiler warning.

Initial implementation of Linux-os-arm.c, with splitting out of x86
stuff from Linux-os.c to Linux-os-x86.c.

Move emit-not-implemented macrolet outside assemble.

x86 builds again.

See issue #36.

Add include files.

Fix typo.

Define reg_PC.

Need to include assert.h and forgot to remove an #if 0.

Actually remove the code that was moved to Linux-os-x86.c.

Remove unneeded #if's.

Smash arm_linux case into x86_linux case.

Add LISPPTR to print out lisp objects as we want.

Use LISPPTR as needed.

Remove blank line.

Merged in arm-rtoy-issue-36 (pull request #67)

Move emit-not-implemented macrolet outside assemble.
Fixes issue #36.
Remove CAT macro; remove parens for NREGS.

Address review comments.

Also added RESEVERED_SPACE to leave some room at the end of each space
that is not mapped so we can detect when we've reached the end of the
space.

Move arm_common into arm_linux and remove Config.arm_common.

Add TODO to reverse this someday if needed.

Stub out routines needed arm-arch.c.

Merged in arm-rtoy-issue-30 (pull request #63)

Support ARM/linux config for create-target.sh
Fixes issue #30
Remove duplicated lines.

Simplify according to review.

Remove UNDEFSYMPATTERN.

Clean up according to review.

Merged in arm-rtoy-issue-29 (pull request #64)

Add Config files for ARM/linux.
Fixes issue #29
Merge branch 'master' into arm-rtoy-issue-34

Mostly revert to master version and use %lx instead of %x.

Remove LISPPTR.

Define and use KB.

Merge branch 'master' into arm-rtoy-issue-31

Can't undef these; they're needed in lisp.c at least.

Merge branch 'master' into arm-rtoy-issue-32

Conflicts:
	src/lisp/Config.arm_linux

Put everything back into Linux-os.c, with arm updates.

Don't have Linux-os-x86.c anymore.

Remove Linux-os-arm.c and Linux-os-x86.c.

Fix typo introduced when moving the x86-64 version of
os_sigcontext_reg to the x86 implementation.

Implement LDM/STM instructions.

See isseu #11.

The syntax is not great but to keep it inline with other instructions
where the condition code is at the end, the list of registers is
really a list.

Some simple tests added in test-disassem.lisp.

Fix typo, and remove extra space.

Merged in arm-rtoy-issue-11 (pull request #70)

Implement LDM/STM instructions.

Fixes issue #11.
Fix up according to review.

Control stack grows up.
Use os_sigcontext_reg instead of os_sigcontext_pc (which should not be
implemented in Linux-os.c for arm).

Remove RESERVED_SPACE stuff. Other ports don't need this.

Clean up according to review.

arm-lispregs.h:
* Add comment for last #endif

arm-validate.h:
* Oops. Forgot to remove one use of RESERVED_SPACE.
* Add comment for last #endif

Merged in arm-rtoy-issue-31 (pull request #65)

Add arm-lispregs.h and arm-validate.h

Fix issue #31.
Fix up according to review.

* Remove <inttypes.h>
* Make gc_abort() call abort().

Merge branch 'master' into arm-rtoy-issue-34

Refactor according to review.

Use abort() when needed
Push #ifdef's up and add #error as needed.

Merge branch 'master' into arm-rtoy-issue-32

Clean up personality() parts, add abort() for unimplemented ARM
version of sigsegv_handler().

Fix up DPRINTF so it should compile.

Reimplement NOT-IMPLEMENTED to dump the symbol name into the code
stream.

See issue #41.  The symbol name is now saved into the instruction
stream as a sequence of bytes.  The length is encoded into a branch
instruction before the bytes so that if the trap is continued, the
bytes are skipped over nicely.

Had to update how EMIT-NOT-IMPLEMENTED was calling NOT-IMPLEMENTED and
a few other places where bare NOT-IMPLEMENTED was called directly.

Abort instead of returning NULL.  Include stdlib.h to get abort
defined.

Tabify the x86_64 and arm os_sigcontext_reg routines to match existing
style.

Merged in arm-rtoy-issue-32 (pull request #68)

Initial implementation of Linux-os.c for ARM
Fixes issue #32.
Remove gc_assert.

Merged in arm-rtoy-issue-34 (pull request #66)

Fix up printf warnings in gc.c.
Fix issue #34.
Update according to review.

o Use do {} while in NOT_IMPLEMENTED and call abort too.
o Return NULL in arch_init.
o Remove some return statements that can't be reached due to
  NOT_IMPLEMENTED.
o Rip out contents of sigill_handler and replace with NOT_IMPLEMENTED.
o Lisp stack grows down for the arm port, so decrement the pointer,
  not increment.

Merge branch 'master' into arm-rtoy-issue-37

Remove trailing space after \ continuation.

Remove extra space.

Merged in arm-rtoy-issue-37 (pull request #69)

Stub out routines needed in arm-arch.c.
Fixes issue #37.
Add support for arm/linux for EXTERN.

(This seems like the wrong place for that!)

Merge branch 'master' into arm-rtoy-issue-43

Merge branch 'master' into arm-rtoy-issue-41

Address review comments.

o Use a B instruction
o Remove unused defregtn a0.

Use ALLOCATION_POINTER and BINDING_STACK_POINTER for C runtime.

Arm uses the static symbols *allocation-pointer* and
*binding-stack-pointer* to implement the allocation pointer and
binding stack pointer instead of dedicated registers.  Update the C
code to use these appropriately.

See issue #44.

Remove some temporarily #if 0'd code.

Stub out restore_fpu for arm.

See issue #45.

Add not-implemented trap code for use by NOT-IMPLEMENTED.

See issue #46.

Initialize current_control_stack_pointer correctly for ARM.

The ARM Lisp stack will grow down, not up like other RISC ports.

See issue #51.

Disable optimization.

We don't need full optimization at this stage of development, so use
-Og instead of -O2.

Merged in arm-rtoy-issue-45 (pull request #73)

Fix issue #45: Stub out restore_fpu for arm.
Merged in arm-rtoy-issue-51 (pull request #74)

Fix issue #51: Initialize current_control_stack_pointer correctly for ARM.
Simplify not-implemented.

Implement emit-relative-branch correctly for ARM.

The offset in the branch imeediate instruction is offset by 8 bytes.
If the instruction is at address n and the offset value in the
instruction is x, the actual target of the branch is the address 
n + x + 8.  This needs to be accounted for when we emit relative
branchs and also in the disassembly of relative branches.

See issue #52.

Fix typo.

Update according to review.

Merged in arm-rtoy-compile-for-debug (pull request #75)

Disable optimization for now
Update according to review.

Fix up comment for relative-branch-offset

Don't need an .extern for arm/linux.

Also put back original indentation of EXTERN.

Add EMIT-NOT-IMPLEMENTED to vops that may already have
implementations.

See issue #53.

Merge branch 'arm-rtoy-issue-41' into arm-rtoy-issue-46-on-41

Conflicts:
	src/compiler/arm/insts.lisp

Put args on the stack in the right place in funcall0 and friends.

Export not-implemented-trap and make sure object-not-list-trap value
doesn't overlap the other trap values.

Implement initial sigill handler for not-implemented trap.

See issue #57.

Merge branch 'master' into arm-rtoy-issue-43

Revert change.

First cut at call_into_lisp.

This implements most of the first part of call_into_lisp to get into
lisp.  The return from lisp is not yet implemented.

See issue #43.

Fix mistake in computing the actual function address.

Use 16 double-float registers.

Implement XEP-ALLOCATE-FRAME

Merged in arm-rtoy-issue-41 (pull request #71)

Reimplement NOT-IMPLEMENTED to dump the symbol name into the code stream.

Fix issue #41.
Merged in arm-rtoy-issue-53 (pull request #77)

Fix issue #53: Add EMIT-NOT-IMPLEMENTED to vops that may already have been implemented
Merged in arm-rtoy-issue-55 (pull request #79)

Fix issue #55: Put args on the stack in the right place in funcall0 and friends.
Merged in arm-rtoy-issue-46-on-41 (pull request #78)

Fix issue #46: NOT-IMPLEMENTED gets its own trap code
Merge branch 'master' into arm-rtoy-issue-56

Kill :vfpv3-d16 and fix float-registers to 16 in vm.lisp.

Update according to review.

Update according to review:

o Kill clear-memory-p
o Use string for not-implemented.

Update according to review.

Fix a few compiler notes.

Merge branch 'master' into arm-rtoy-issue-57

Merge branch 'master' into arm-rtoy-issue-43

Merge branch 'arm-rtoy-issue-44' into arm-rtoy-issue-43

Update according to review.

o Remove .text from FUNCDEF
o The header word for undefined_tramp and closure_tramp is wrong.  ARM
  is little-endian so the type_FunctionHeader byte needs to come
  first.
o Don't need the .align for undefined_tramp and closure_tramp.
o Align function_end_breakpoint_guts to 8 bytes because sparc
  does. Not sure if this is really necessary, though.

Fix typo in getting the high part of the address of
current_control_stack_pointer.

Use inst byte instead of emit-byte

See issue #58.

Fix typo.

Implement undefineds-assem.S.

See issue #35.

Be lenient in disassembling CMP

When disassembling a CMP instruction, allow the dst register field of
the instruction to be any value instead of fixing it to 0. The dst
isn't printed anyway.

See issue #54.

Better disassembly for MOV Rd, Rs

The actual instruction is LSL Rd, Rs, #0, which we'd like disassembled
as MOV, which is a bit easier to understand.

See issue #15.

Add test from issue #54.

Include the name when disassembling UDF.

See issue #60.

When disassembling the UDF instruction, print out the name as a
disassembly note.

We also updated the udf printer to print the immediate using the
current radix instead of always using base 10.  This for consistency
with how other numbers are printed.

Access static symbols correctly, and save the current C stack pointer
and frame pointer to the static symbols so Lisp can access them.

Swap CFP and OCFP register assignments.

r11 is preserved across C function calls, so it's advantageous to make
CFP be r11 and OCFP be r12.  r12 is not preserved across call, but
OCFP is stored on the Lisp stack, so we can easily recover the value.

See issue #62.

Merged in arm-rtoy-issue-58 (pull request #84)

Fix issue #58: Use inst byte instead of emit-byte
Merged in arm-rtoy-issue-44 (pull request #72)

Fix issue 44: Use ALLOCATION_POINTER and BINDING_STACK_POINTER in C runtime
Put the printer def all on one line, per review comments.

Merged in arm-rtoy-issue-15 (pull request #87)

Fix issue #15: Better disassembly for MOV Rd, Rs
Update according to review.

Add new function-header trap value and use it in XEP-ALLOCATE-FRAME.

Merged in arm-rtoy-issue-54a (pull request #86)

Fix issue #54: Be lenient in disassembling CMP
Merged in arm-rtoy-issue-56 (pull request #82)

Fix issue #56: Use 16 double-float registers.
Implement BX instruction and fix %LI.

The BX instruction is not previously implemented and we need it.

We also take this opportunity to correct the %LI function.
Previously, the MOVW instruction would be used after the MOVT
instruction, but that would clear the high bits set by MOVT.

See issue #63.

Merged in arm-rtoy-issue-63 (pull request #90)

Fix issue #63: Implement BX instruction and fix %LI.
Merge branch 'master' into arm-rtoy-issue-43

Add missing NOT-IMPLEMENTED stuff.

Add more missing NOT-IMPLEMENTED stuff.

Update according to review.

* Remove FUNCDEF
* Upper case load/store/load_symbol_value/store_symbol_value.
* Use ldr reg, =foo instead of movw/movt.
* Insert spaces around commas.
* Rename SET_SIZE to ENDFUNC.
* Change .global to .globl.
* Remove unneeded .text and .align.

Store the function pointer into the fdefn-raw-addr-slot instead of the
raw address.

This matches sparc and ppc and slightly simplifies scavenging the
fdefn object.

Need to load up arm assembly files.

Merged in arm-rtoy-issue-66 (pull request #93)

Fix issue #66: Need to load up arm assembly files.
Merged in arm-rtoy-issue-65 (pull request #92)

Fix issue #65: Store function pointer into fdefn-raw-addr-slot instead of the raw address
Merged in arm-rtoy-issue-48 (pull request #83)

Fix issue #48: Implement XEP-ALLOCATE-FRAME
Merged in arm-rtoy-issue-52 (pull request #76)

Fix issue #52: Implement emit-relative-branch correctly for ARM.
Merge branch 'master' into arm-rtoy-issue-43

Add D(x)=x, for completeness.

Remove FIXME comment and clean up following comment.

Cuddle brace with case statement.

Renumber traps so we don't use 16 as a trap value.

Also took opportunity to make add pseudo-atomic trap as part of the
defenum.

Fix some comments.

Remove LOAD/STORE and add .pool to ENDFUNC.

Merged in arm-rtoy-issue-57 (pull request #80)

Fix issue #57: Implement initial sigill handler for not-implemented trap.
Make D(foo) do nothing and add TODO to clean this up.

Remove ambiguous part of comment.

Merged in arm-rtoy-issue-35 (pull request #85)

Fix issue #35: Implement undefineds-assem.S.
Remove bogus comment.

Merged in arm-rtoy-issue-43 (pull request #81)

Fix issue #43: Initial implemention of call_into_lisp
Merged in arm-rtoy-issue-67 (pull request #94)

Fix issue #67: Renumber traps so we don't use 16 as a trap value.
Change NOT_IMPLEMENTED to require a message and value.  Useful for
printing additional information.

Implement call_into_c.

Remove linkage-tables feature for arm/linux

The initial port for arm will not use linkage-tables.  Update code
accordingly.

tools/cross-scripts/cross-x86-arm.lisp:
o Remove :linkage-table from the set of features we want for arm.  Put
  it on the mis-features list.

lisp/undefineds.h:
o Update with symbols arm/linux wants.

compiler/generic/new-genesis.lisp:
o Update symbol lookup for our arm implementation that doesn't use the
  PVE_stub stuff.
o Add implementation for arm for handling fixups
o Minor fix adding conditionalization for linkage-table that was
  inadverntly missed long ago.

More changes to support building without linkage tables.

Config.arm_linux:
o Update NM to use linux-nm.
o Need -lutil.

linux-nm:
o Update to linux-nm to replace stub_ so we use the correct addresses
  in genesis.

Remove unneeded prints.

The new symbols are needed by linux and not just arm and linux.

Merged in arm-rtoy-issue-64 (pull request #91)

Fix issue #64: Add missing NOT-IMPLEMENTED stuff.
Fix typos and bugs in fixups.

o Encoding for the movt inst was using the encoding for movw.
o Fix typo: It's :movw, not :mov3!
o Reorder fixup code to use movw/movt, not movt/movw which wipes out
  the top 16 bits set by movt.

Update according to review.

Merged in arm-rtoy-issue-61 (pull request #97)

Fix issue #61: Remove linkage-tables feature for arm/linux
Merge branch 'master' into arm-rtoy-issue-60

Fix call to snarf-not-implemented-name.

Add -X option to (break) after writing the core file.

Useful to stop the cross-compilation just before it quits so we can
run compile-file or other tests using the cross-compiled target code.

Print immediate values in the current output radix.

Oops.  Revert the print-immed change.

Print immediate values in the current output radix.

Merged in arm-rtoy-issue-69 (pull request #98)

Fix issue #69: Add -X option to (break) after writing the core file.
Merged in arm-rtoy-issue-59 (pull request #96)

Fix issue #59: Implement call_into_c.
Use FORMAT to print out immediate values.

Merge branch 'master' of gitlab cmucl (snapshot-2015-07) into arm-rtoy-sync-2015-07

Merged in arm-rtoy-sync-2015-07 (pull request #100)

Sync to upstream snapshot-2015-07

Merge branch 'master' into arm-rtoy-issue-70

Merge branch 'master' into arm-rtoy-issue-68

Merge branch 'master' into arm-rtoy-issue-62

Attempted fix for call_into_lisp():
  * modified call_into_lisp() to set the current frame pointer to the current stack pointer upon entry into the subroutine
  * cleaned up use of reg_NL1 as base pointer to passed in arguments

Attempted fix for arm-lispregs.h defintions
  * defined all register names as part of REGNAMES
  * corrected typo where reg_CSP was referred to as reg_SP

removing extra spaces and indentation

Merged in arm-rtoy-issue-70 (pull request #99)

Fix issue #70: Print immediate values in the current output radix.

Merge branch 'master' into arm-rtoy-issue-68

Merge branch 'master' into arm-rtoy-issue-60

Merge tag 'snapshot-2015-09' of /Volumes/share2/src/clnet/cmucl/dev/cmucl into arm-rtoy-sync-2015-09

Snapshot 2015-09

Merged in arm-rtoy-sync-2015-09 (pull request #102)

Sync to upstream snapshot-2015-09

Merged in arm-rtoy-issue-60 (pull request #88)

Fix issue #60: Include the name when disassembling UDF.

reverting to original behavior with reg_CFP set to reg_NL0

Merge branch 'master' into arm-rtoy-issue-68

New version of NOT_IMPLEMENTED, suggested in review.

Rename NOT_IMPLEMENTED_MESSAGE to NOT_IMPLEMENTED and use it
everywhere.

Merge branch 'master' into arm-rtoy-issue-62

Remove null statement.

Fix issue #71: Typo introduced in pull request #99.

Add the closing paren for eval-when back.

Merged in arm-rtoy-issue-71 (pull request #103)

Fix issue #71: Typo introduced in pull request #99.

Merge branch 'master' into arm-rtoy-issue-68

Use lose() instead of NOT_IMPLEMENTED.

Merged in arm-rtoy-issue-68 (pull request #95)

Fix issue #68: Change NOT_IMPLEMENTED to require a message and value.  Useful for

corrected typo -- NL1 is the pointer to the control-stack, not NL0

Initialize Lisp control stack pointer correctly.

Fixes #72.  Set the control stack pointer a few words before the end
of the control stack space so that the pointer points to an area that
is actually writable.

Merge branch 'master' into arm-rtoy-issue-62

Undo changes.

Restore saved current_control_frame_pointer after returning.

Implement PRINT vop

Back up just one word to get to the writable area.

Stack grows down so use negative offsets for load/store stack-tn.

Control stack offsets are negative.

clarifying that the arguments are relative to the current frame pointer

Adjust the stack pointer and push the args correctly for the downward
growing control stack on ARM.

Use LIP as the temp register for restoring CSP and CFP.

Revert change.

Store args in the correct order for the downward growing stack in
funcall0 and friends.

Merged in arm-rtoy-issue-62 (pull request #89)

Restore CFP after calling out to C

Merge branch 'master' into arm-rtoy-issue-72

Simplify the macros load-stack-tn and store-stack-tn.

Remove some trailing blanks at eol.

Merged in arm-rtoy-issue-72 (pull request #104)

Initialize Lisp control stack pointer correctly.

Fix issue #72

Merge branch 'master' into arm-rtoy-issue-50-print

Fix obvious typo introduced in pull request #89.

We changed the register to hold the addresses, but weren't using it to
load the values from that address!

Merged in arm-rtoy-pull-request-89-correction (pull request #106)

Fix obvious typo introduced in pull request #89.

Update lisp stack offsets in call_into_c.

The lisp stack grows down so the offsets are negative!

Merge branch 'master' into arm-rtoy-issue-50-print

Only allocate 5 words.

The comment is updated as to what happens. I've verified this in the
debugger and for each case I've looked at, the C stack pointer is
double-word aligned just before the branch to the actual C routine.

Merged in arm-jboone-asm-2015-09 (pull request #101)

correction of arm-lispregs.h definitions

Merge branch 'master' into arm-rtoy-issue-50-print

Document the expected stack layout for call_into_lisp.

Add a few more comments.

* Remove some comments that should be in call_into_c.
* Zero out the unused arg stack slots, just for debugging purposes.

Merged in arm-rtoy-issue-50-print (pull request #105)

Implement PRINT vop.  Fixes issue #50.

Merge tag 'snapshot-2016-06' of common-lisp.net:cmucl/cmucl into rtoy-sync-snapshot-2016-06

Snapshot 2016-06

Merged in rtoy-sync-snapshot-2016-06 (pull request #107)

Sync upstream snapshot-2016-06

Define SIGNAL_STACK_SIZE

After merging the upstream snapshot, SIGNAL_STACK_SIZE needs to be
defined.  Define it so we can build lisp again.

Merge branch 'master' of common-lisp.net:cmucl/cmucl into rtoy-sync-21b

Merged in rtoy-sync-21b (pull request #108)

Sync to 21b

Use altstack on ARM

When installing a low-level handler, use an alternate stack for
handling signals.

This seems to allow our handlers to handle signals again. Recently, we
started getting faults apparently before our handlers ever got a
chance to run.  We can make progress now when running kernel.core.

Long ago, this all worked, but I don't know what made it stop.  For
one, the OS on my panda board was updated (and downgraded).  Don't
know what made it stop working.

Debug prints in sigill handler.

Merge tag '21c' into arm-rtoy-sync-21c

Release 21c

Sync to upstream 21c

Set address of where the lisp control stack ends

Must of been deleted at some point when merging with upstream.

* added new conditional around F(sigvec) that excludes it from __linux__ as well

* corrected typo (changed __FReeBSD__ to __FreeBSD__)

Add some Gray stream extensions

Merge branch 'gray' into 'master'

Add some Gray stream extensions

See merge request cmucl/cmucl!181
Add support for Gray interactive-stream-p

Merge branch 'gray-interactive' into 'master'

Add support for Gray interactive-stream-p

See merge request cmucl/cmucl!183
Fix #260:  Mention hemlock must loaded for -edit/-slave to exist

Merge branch 'issue-260-hemlock-in-core-for-options' into 'master'

Fix #260:  Mention hemlock must loaded for -edit/-slave to exist

Closes #260

See merge request cmucl/cmucl!182
Reorder docstring for defpackage to match CLHS

Minor tweak to list the options to `defpackage` to match the order
showns in the CLHS entry for `defpackage`.  Since it's just a change
in the docstring, no changes in behavior anywhere.

Fix #271: Update ASDF to version 3.3.7

Merge branch 'issue-271-update-asdf-3.3.7' into 'master'

Fix #271: Update ASDF to version 3.3.7

See merge request cmucl/cmucl!186
Fix #256: loop for var nil works

Merge branch 'issue-256-loop-var-nil' into 'master'

Fix #256: loop for var nil works

Closes #256

See merge request cmucl/cmucl!185
Fix #272;  Move scavenge code for static vectors to its own function

Merge branch 'issue-272-add-scav-static-vector-fcn' into 'master'

Fix #272;  Move scavenge code for static vectors to its own function

Closes #272

See merge request cmucl/cmucl!187
Update with recent bug fixes

Forgot to update this when each bug was fixed.

Fix #278: Add more debugging prints for gencgc

Merge branch 'issue-278-more-gencgc-debug-prints' into 'master'

Fix #278: Add more debugging prints for gencgc

Closes #278

See merge request cmucl/cmucl!193
Fix #277: return least-positive float for rationals close to it

Merge branch 'issue-277-float-ratio-float-least-positive-float' into 'master'

Fix #277: return least-positive float for rationals close to it

Closes #277

See merge request cmucl/cmucl!191
Fix #274: 1d999999 hangs cmucl

Merge branch 'issue-274-make-float-of-huge-numbers' into 'master'

Fix #274: 1d999999 hangs cmucl

Closes #274

See merge request cmucl/cmucl!189
Fix #276: Implement xoroshiro128**

Merge branch 'issue-276-xoroshiro128starstar' into 'master'

Fix #276: Implement xoroshiro128**

Closes #276

See merge request cmucl/cmucl!192
Update notes with more fixed issue numbers

Forgot to update the release notes when we fixed various issues.

Fix #283: Add VOP for integer-length for an (unsigned-byte 32) arg

Merge branch 'issue-283-unsigned-integer-length-vop' into 'master'

Fix #283: Add VOP for integer-length for an (unsigned-byte 32) arg

Closes #283

See merge request cmucl/cmucl!194
Update release notes for #283

Forgot to update release notes after fixing #283.

Checkout ansi-tests cmucl-expected-failures again

Issue https://gitlab.common-lisp.net/cmucl/cmucl/-/issues/276 has
landed and we've merged the ansi-test branch issue-276-xoroshiro to
cmucl-expected-failures.  Thus, we can check out
cmucl-expected-failures again for running ansi-tests.

Fix #284:  Optimize signed-byte-32-int-len VOP

Take advantage of the fact that `BSR` will not modify the destination
register if the source is 0.  We already use this idea for the
`unsigne-byte-32-int-len` VOP, so let's do the same for this VOP.  It
saves an instruction and a branch.

Update release notes for #284

Remove comment to rtfm for signed-byte-32-len

Not sure what the comment is referring too.  It came in the huge
commit [c0c98ded6c].  I think the implementation is correct.

Optimize unsigned-byte-32-int-len some more

Take advantage of the fact that BSR actually moves the src to the dst
register when the src is 0.  The Intel docs don't says this, but gcc,
LLVM, and MSVC compilers basically assume this when generating code
that uses BSR.

Add some tests for integer-length of signed and unsigned 32-bit ints

Define functions that use the VOPs for integer-length of signed and
unsigned 32-bit integers.

Test a few "interesting" integers with these functions.

Merge branch 'issue-284-optimize-signed-byte-32-int-len-vop' into 'master'

Fix #284:  Optimize signed and unsigned 32-bit integer length VOPs

Closes #284

See merge request cmucl/cmucl!195
Fix #287: Clean up make-float-aux

Merge branch 'issue-287-clean-up-make-float-aux' into 'master'

Fix #287: Clean up make-float-aux

Closes #287

See merge request cmucl/cmucl!196
Fix #290:  Neatly pprint with-float-traps-masked and friends

Merge branch 'issue-290-pprint-with-float-traps' into 'master'

Fix #290:  Neatly pprint with-float-traps-masked and friends

Closes #290

See merge request cmucl/cmucl!199
Fix #288: Re-enable deftransform for random integers

Merge branch 'issue-288-re-enable-deftransform-random-int' into 'master'

Fix #288: Re-enable deftransform for random integers

Closes #288

See merge request cmucl/cmucl!198
Fix #291:  pprint handler-case neatly

Merge branch 'issue-291-pprint-handler-case' into 'master'

Fix #291:  pprint handler-case neatly

Closes #291

See merge request cmucl/cmucl!200
Update notes with fixed issues

Forgot to add #288, #290, and #291 to the list of fixed issues.

Fix #298: Add with-float-rounding-mode macro

Add `ext:with-float-rounding-mode` macro to set the FP rounding mode
to be used when executing the body.

Update cmucl.pot with new docstrings

Add tests for with-float-rounding-mode

Not 100% sure the expected results are right.  They seem plausible,
but I didn't actually analyze that the rounding is correct.

Merge tag 'snapshot-2018-03' of common-lisp.net:cmucl/cmucl into arm-rtoy-sync-2018-03

Snapshot 2018-03
2018-03-02

Merged in arm-rtoy-sync-2018-03 (pull request #111)

Sync to upstream 2018-03 snapshot

Fix up textdomains

Needed to add arm domain for with-textdomain because it was missing.

Define COMPUTE-LRA-FROM-CODE

Copied from sparc.

We need this instruction to implement DEFINE-FULL-CALL for CALL-NAMED
(at least).

Add missing closing paren

Not sure how this was deleted, but I didn't bother to figure it out.

Fix typo in loadw/storew.

We had (typep '(signed-byte 13) offs). Duh! It should be
(typep offs '(signed-byte 13)).

Merged in linux-jboone-remove-sigvec (pull request #110)

* added new conditional around F(sigvec) that excludes it from __linux__ as well

Approved-by: Raymond Toy <toy.raymond@gmail.com>

Merge branch 'master' into arm-rtoy-issue-76

Print out more info for unknown udf code.

Reorder registers again.

Swap nfp with ocfp because r1 (nfp) is an output register and that
conflicts with the wired-tn nfp.  I don't think ocfp is wired to
anything. 

Update C and assembly code accordingly.

Merged master, fixing conflicts

Remove unneeded nop

ARM doesn't have a branch-delay slot.

Merge branch 'arm-rtoy-issue-76-rename-regs-2' into arm-rtoy-issue-76

Add support for trapping error and cerror

Recognize the error and cerror traps and handle them.  Not yet really
tested, but looks plausible based on other implementations.

Fix #74: Implement symbol-value

Basically copied the sparc version.

With this change, we can see that we're getting the symbol value of
the symbol `*LISP-INITIALIZATION-FUNCTIONS*`.  The value is:

```
(gdb) call print(0x41b103ab)
$1=	  0x41b103ab: list pointer
$2=	    car: 0x41b0fae9: #<ptr to function header>
$3=	    cdr: 0x41af476b: ($4=#<ptr to function header> $5=#<ptr to function header> $6=#<ptr to function header> $7=#<ptr to function header> ...)
```

Fix up call-named

Put in emit-not-implemented so we can tell and compute the target
address correctly because we forgot to handle the tag bits.

Implement arch_internal_error_arguments

Based on sparc and ppc version where we just grab the address of the
word after the PC.

Implement allocate-full-call-frame

Basically copied the sparc version.

Merged in arm-rtoy-issue-74 (pull request #112)

Fix #74: Implement symbol-value

Approved-by: Raymond Toy <toy.raymond@gmail.com>

Remove old attempts at call-named.

Just leave the sparc and RT versions; with the RT version being used.

Rename and add new reg assignments like RT

This is a first step at implementing call-named (as mentioned in issue
#76).  

As best as I can tell it would be really useful to have nfp and cname
as actual registers, which the RT port does.  Thus, remove nl1 and nl2
and add nfp and cname.  We didn't lose 2 of our non-descriptors
though. The RT port makes nfp and ocfp non-descriptors so we recover
that.  We'll see if this works out ok.

So, rename and add the registers, updating the list of non-descriptors
and descriptors.

Update arm-assem.s with these changes.  Note that we make nargs be r2,
the third arg to call_into_lisp. This slightly easier to read the
assembly code.

Update arm-lispregs.h to have the right register assignments and
names.

With this change running the kernel core gets as far as it used to
with the ame error message.

Merged in arm-rtoy-issue-75 (pull request #113)

Fix #75: Implement allocate-full-call-frame

Approved-by: Raymond Toy <toy.raymond@gmail.com>

Undo change to register-arg-names

We still have 3 args for this port.

Merged in arm-rtoy-issue-76-rename-regs (pull request #114)

Rename and add new reg assignments like RT

Fix #76: Implement call-named

* Define COMPUTE-LRA-FROM-CODE

    Copied from sparc.

    We need this instruction to implement DEFINE-FULL-CALL for CALL-NAMED
    (at least).

* Add missing closing paren

    Not sure how this was deleted, but I didn't bother to figure it out.

* Fix typo in loadw/storew.

    We had (typep '(signed-byte 13) offs). Duh! It should be
    (typep offs '(signed-byte 13)).

* Merged master, fixing conflicts

* Reorder registers again.

    Swap nfp with ocfp because r1 (nfp) is an output register and that
    conflicts with the wired-tn nfp.  I don't think ocfp is wired to
    anything.

    Update C and assembly code accordingly.

* Fix up call-named

    Put in emit-not-implemented so we can tell and compute the target
    address correctly because we forgot to handle the tag bits.

* Update arith.lisp because allow one arg to be on the unsigned
  stack.  (Not really sure this is needed, but the RT port does this.)

* Make NFP hard-wired like the RT port (and all the other RISC ports).

Reorder bug list to be numerically ascending.

No other changes.

Fix #297:  Print new-assem:assemble with less indentation

Merge branch 'issue-297-pprint-assemble' into 'master'

Fix #297:  Print new-assem:assemble with less indentation

Closes #297

See merge request cmucl/cmucl!203
Fix #300: Reduce code duplication in random

Merge branch 'issue-300-reduce-code-dup-in-random' into 'master'

Fix #300: Reduce code duplication in random

Closes #300

See merge request cmucl/cmucl!206
Make variable names consistent.
Fix #295:  Add docstring for define-assembly-routine

Merge branch 'issue-295-docstring-for-define-assembly-routine' into 'master'

Fix #295:  Add docstring for define-assembly-routine

Closes #295

See merge request cmucl/cmucl!204
Fix typos caused by renaming

We renamed `old-rounding-mode` to `old-mode`, but forgot to update all
uses.

Fix up docstring for with-float-rounding-mode

Mention that the allowed values are the same as the values for the
rounding-mode in `set-floating-point-modes`.

Change indentation of docstring slightly so that it prints nicely via
`describe`.  (Emacs doesn't indent the docstring by the right amount
for `describe`.)

Remove extra trailing space after period.

Fix #294: Implement assembly routine for xoroshiro update

Merge branch 'issue-294-xoroshiro-lisp-assem-routine' into 'master'

Fix #294: Implement assembly routine for xoroshiro update

Closes #294

See merge request cmucl/cmucl!202
Update cmucl.pot for changed docstring for with-float-rounding-mode.

Merge branch 'issue-298-with-float-rounding-mode' into 'master'

Fix #298: Add with-float-rounding-mode macro

Closes #298

See merge request cmucl/cmucl!205
Fix #299: Use xoroshiro assembly routine for x86

Merge branch 'issue-299-enable-xoroshiro-assem-routine' into 'master'

Fix #299: Use xoroshiro assembly routine for x86

Closes #299, #295, #300, #297, and #294

See merge request cmucl/cmucl!208
Add comment about using -V and use a better version

Explain why we use `-V` for `bin/make-dist.sh`.  And instead of using
"ci-build" as the version, use `git describe --dirty` as the version.
This lets us know better what version was actually used for the build.

Fix #296:  Fix disassembly of MOVD instruction

Merge branch 'issue-296-fix-movd-disassembly' into 'master'

Fix #296:  Fix disassembly of MOVD instruction

Closes #296 and #300

See merge request cmucl/cmucl!207
Fix #275:  FP underflow in reader allows restart with 0

Merge branch 'issue-275b-signal-float-underflow' into 'master'

Fix #275:  FP underflow in reader allows restart with 0

Closes #275 and #287

See merge request cmucl/cmucl!197
Fix #293:  Allow restarts for FP overflow in reader

Merge branch 'issue-293-restart-on-reader-fp-overflow' into 'master'

Fix #293:  Allow restarts for FP overflow in reader

Closes #293, #275, and #287

See merge request cmucl/cmucl!201
Update with fixed tickets that we forgot to add earlier.

Use snapshot-2024-04 to run the CI

Snapshots have been uploaded so use them for running the CI.

Remove // comment

According to [fc0bd2a6f0] this change was supposed to be removed but
never was.  Let's do it now.  This was also the only place where we
used C++ comment-style.

Fix #303: Remove unused var *assert-not-standard-readtable*

Merge branch 'issue-303-remove-unused-assert-not-standard-readtable' into 'master'

Fix #303: Remove unused var *assert-not-standard-readtable*

Closes #303

See merge request cmucl/cmucl!209
Fix a double free detected by the GCC analyzer

The obuffer variable contained a pointer to a block of memory freed by
realloc(3) if the ERANGE case was executed more than once.  Afterward,
if the 0 case executed, obuffer would be passed to free(3) causing a
double free.

This change sets the value of obuffer to buffer immediately after a
successful call to realloc(3) ensure it is always NULL or a valid
block of memory.

Merge branch 'file-author' into 'master'

Fix a double free detected by the GCC analyzer

See merge request cmucl/cmucl!210
Fix an incorrect error check detected by the GCC analyzer

The return value of open(2) should be compared with -1 to check for an
error but it was compared with 0 instead.

Fix the retry limit for resizing password entry buffer

The previous version checked that the buffer had not exceeded the
limit of 1MiB after the buffer was resized.  In otherwords, the buffer
is allowed to grow past the limit before the retry loop is exited.

This change moves the check of the buffer size before the allocation
occurs.  Now, the retry loop is exited before the buffer would grow
past the limit.

Merge branch 'open-error' into 'master'

Fix an incorrect error check detected by the GCC analyzer

See merge request cmucl/cmucl!211
Merge branch 'retry-limit' into 'master'

Fix the retry limit for resizing password entry buffer

See merge request cmucl/cmucl!212
Remove the definition of __NO_CTYPE from the Linux build

The definition of this macro disables the definition of isalpha(3) and
other symbols from ctype.h from being defined as macros that depend on
internal glibc functions not present in all versions of glibc.

At the time, that allowed binaries compiled against newer versions of
glibc to run on systems with older versions of glibc.  However, today
the internal functions have been available for more than two decades
and so this feature does not appreciably affect compatibility.

Merge branch 'no-ctype' into 'master'

Remove the definition of __NO_CTYPE from the Linux build

See merge request cmucl/cmucl!214
Return an error if a copy of *CMUCL-LIB* cannot be allocated

Previously, the malloc and strdup calls were assumed to succeed.  Now,
their return values are checked for an error signal.  Consistent with
other failures in this file, a value of -1 is returned.

Merge branch 'elf-c-unchecked-malloc' into 'master'

Return an error if a copy of *CMUCL-LIB* cannot be allocated

See merge request cmucl/cmucl!217
Fix #314: tanh incorrect for large args

Merge branch 'issue-314-complex-tanh-incorrect-for-large-arg' into 'master'

Fix #314: tanh incorrect for large args

Closes #314

See merge request cmucl/cmucl!219
Update release notes for #314

Forgot to update the release notes when working on #314.

Merge tag 'snapshot-2024-04' into arm-rtoy-sync-snapshot-2024-04

Snapshot 2024-04

Merge remote-tracking branch 'bitbucket/arm-rtoy-sync-snapshot-2024-04'

Merge branch 'master' into issue-156-take-2-nan-comparison

Handle NaN in ir1-transform->=

Previously, when `x` and `y` are the same leaf, the transform would
immediately return `T`.  However, this is not correct if `x` is a
float because NaN >= NaN is always false.  So add check that `x` is
not a float.

Manually tested (for now) the with
```
(defun bar (x)
  (declare (double-float x))
  (<= x y))
```
returns `NIL` for `(bar *nan*)` and `T` for `(bar pi)`.  Without this,
it always returned `T`.

Update cmucl.pot too because it was regenerated.

Raymond Toy pushed to branch issue-156-take-2-nan-comparison at cmucl / cmucl

Commits:

30 changed files:

The diff was not included because it is too large.