#124: Failure to be an SBCL build host ------------------------+--------------------------------------------------- Reporter: ehuelsmann | Owner: nobody Type: defect | Status: new Priority: major | Milestone: Component: libraries | Version: Keywords: | ------------------------+--------------------------------------------------- When building SBCL with ABCL, I'm getting the error
#<XP:XP-STRUCTURE ...> isn't of type STREAM
Failing to continue the SBCL build.
The issue here is that our pretty printer structure - which wraps a stream - isn't itself a stream. Somehow, the pretty printer stream wrapper is passed into ABCL internals expecting a STREAM, though.
#124: Failure to be an SBCL build host (XP-STRUCTURE not a STREAM) -------------------------+-------------------------------------------------- Reporter: ehuelsmann | Owner: nobody Type: defect | Status: new Priority: major | Milestone: Component: libraries | Version: Resolution: | Keywords: -------------------------+--------------------------------------------------
#124: Failure to be an SBCL build host (XP-STRUCTURE not a STREAM) ------------------------+--------------------------------------------------- Reporter: ehuelsmann | Owner: nobody Type: defect | Status: new Priority: major | Milestone: Component: libraries | Version: Keywords: | ------------------------+---------------------------------------------------
Comment(by mevenson):
r13529 (post abcl-0.26.2) seems to hang after building sbcl-1.0.50/src/compiler/node.lisp much earlier than the XP structure problem.
#124: Failure to be an SBCL build host (XP-STRUCTURE not a STREAM) ------------------------+--------------------------------------------------- Reporter: ehuelsmann | Owner: nobody Type: defect | Status: new Priority: major | Milestone: Component: libraries | Version: Keywords: | ------------------------+--------------------------------------------------- Changes (by mevenson):
* cc: mevenson (added)
#124: Failure to be an SBCL build host (XP-STRUCTURE not a STREAM) ------------------------+--------------------------------------------------- Reporter: ehuelsmann | Owner: nobody Type: defect | Status: new Priority: major | Milestone: Component: libraries | Version: Keywords: | ------------------------+---------------------------------------------------
Comment(by mevenson):
Replying to [comment:2 mevenson]:
r13529 (post abcl-0.26.2) seems to hang after building
sbcl-1.0.50/src/compiler/node.lisp much earlier than the XP structure problem.
The hang stems from the writing of large constants to separate files within the fasl introduced in r13514.
#124: Failure to be an SBCL build host (XP-STRUCTURE not a STREAM) ------------------------+--------------------------------------------------- Reporter: ehuelsmann | Owner: nobody Type: defect | Status: new Priority: major | Milestone: Component: libraries | Version: Keywords: | ------------------------+---------------------------------------------------
Comment(by mevenson):
Replying to [comment:4 mevenson]:
To reproduce:
1) Download [http://prdownloads.sourceforge.net/sbcl/sbcl-1.0.50-source.tar.bz2?download sbcl-1.0.50-source.tar.bz].
2) Unpack into ~/src/sbcl-1.0.50-source.tar.bz
3) cmd$ cd ~/src/sbcl-1.0.50 && sh make.sh --xc-host="$HOME/work/abcl/abcl --noinit"
#124: Failure to be an SBCL build host (XP-STRUCTURE not a STREAM) ------------------------+--------------------------------------------------- Reporter: ehuelsmann | Owner: nobody Type: defect | Status: new Priority: major | Milestone: Component: libraries | Version: Keywords: | ------------------------+---------------------------------------------------
Comment(by mevenson):
Replying to [comment:4 mevenson]:
The hang stems from the writing of large constants to separate files
within the fasl introduced in r13514.
Not really a "hang" but a really inefficient algorithm in the presence of a file containing a large number of compilands. Addressed in r13531.
#124: Failure to be an SBCL build host (XP-STRUCTURE not a STREAM) ------------------------+--------------------------------------------------- Reporter: ehuelsmann | Owner: nobody Type: defect | Status: new Priority: major | Milestone: Component: libraries | Version: Keywords: | ------------------------+---------------------------------------------------
Comment(by mevenson):
Replying to [comment:7 mevenson]:
Further on, another compiler regression occurs in the compilation of
src/compiler/globaldb.lisp which comes from SYSTEM::PROCESS-TOPLEVEL-MOP .ENSURE-METHOD
Fixed this problem in r13532.
#124: Failure to be an SBCL build host (XP-STRUCTURE not a STREAM) ------------------------+--------------------------------------------------- Reporter: ehuelsmann | Owner: nobody Type: defect | Status: new Priority: major | Milestone: Component: libraries | Version: Keywords: | ------------------------+---------------------------------------------------
Comment(by mevenson):
The cause of the XP error has been isolated into a test case separate from building SBCL and recorded in ticket #165.
#124: Failure to be an SBCL build host ------------------------+--------------------------------------------------- Reporter: ehuelsmann | Owner: mevenson Type: defect | Status: assigned Priority: major | Milestone: Component: libraries | Version: Keywords: | ------------------------+--------------------------------------------------- Changes (by mevenson):
* owner: nobody => mevenson * status: new => assigned
Comment:
The patch attached to ticket #165 gets the build quite a bit further, past the XP error.
But then the build hangs on compiling src/code/unix.lisp in (DEFUN UNIX- SELECT) under x64-solaris-5.11.oi148. But that's a bit of an obscure architecture failing on a UNIX specific portion so this is inconclusive. Need to test on other architectures.
#124: Failure to be an SBCL build host ----------------------------+----------------------------------------------- Reporter: ehuelsmann | Owner: mevenson Type: defect | Status: assigned Priority: major | Milestone: 0.27 Component: other | Version: 0.27 Keywords: sbcl-buildhost | ----------------------------+----------------------------------------------- Changes (by mevenson):
* keywords: => sbcl-buildhost * version: => 0.27 * component: libraries => other * milestone: => 0.27
#124: Failure to be an SBCL build host ----------------------------+----------------------------------------------- Reporter: ehuelsmann | Owner: mevenson Type: defect | Status: assigned Priority: major | Milestone: 0.27 Component: other | Version: 0.27 Keywords: sbcl-buildhost | ----------------------------+-----------------------------------------------
Comment(by mevenson):
With fixing #165, ABCL works as a buildhost for SBCL on OSX again.
Problems remain with x64-solaris-5.11.oi148
#124: Failure to be an SBCL build host -------------------------+-------------------------------------------------- Reporter: ehuelsmann | Owner: mevenson Type: defect | Status: closed Priority: major | Milestone: 0.27 Component: other | Version: 0.27 Resolution: fixed | Keywords: sbcl-buildhost -------------------------+-------------------------------------------------- Changes (by mevenson):
* status: assigned => closed * resolution: => fixed
Comment:
The problem with x64-solaris-5.11.oi148 lies somewhere in the toolchain, as sbcl building itself gets "stuck" here.
Closing.
armedbear-ticket@common-lisp.net