[armedbear-devel] abcl-1.1.0 release status
Overall, I think we are more-or-less ready to release the long awaited abcl-1.1.0 with the exception of resolving the outstanding ANSI regression failures. I have moved all previously outstanding issues to the abcl-1.1.1 milestone, and I have added as three blocker tickets to the abcl-1.1.0 milestone to decide on what to do. http://trac.common-lisp.net/armedbear/ticket/265 http://trac.common-lisp.net/armedbear/ticket/266 http://trac.common-lisp.net/armedbear/ticket/267 For each of these issues, I have identified the changeset that caused the regression via bisecting the source tree. If the result of the bisection is correct, all three issues should be relatively easy to track down, as "not much happened" in the commits. But maybe something was screwy in my process (I did it "on the road" from an iPad ssh client last week in a non-traditional--for me--working environment). Additionally, we should probably check the Novemeber Quicklisp test reports to see if anything has been broken. I seem to see a problem with Hunchentoot under SLIME regarding the PATHNAME reader, but this may be something local to me. I need to finish writing up the CHANGES, and make a quick pass through the manual, updating the grovelled docstring sections. Hopefully, we can have a quick consultation on #abcl tonight at 2100 CST to see if these plans make sense. -- "A screaming comes across the sky. It has happened before but there is nothing to compare to it now."
On Nov 26, 2012, at 15:11, Mark Evenson <evenson@panix.com> wrote:
Overall, I think we are more-or-less ready to release the long awaited abcl-1.1.0 with the exception of resolving the outstanding ANSI regression failures.
Hooray! [...]
Additionally, we should probably check the Novemeber Quicklisp test reports to see if anything has been broken. I seem to see a problem with Hunchentoot under SLIME regarding the PATHNAME reader, but this may be something local to me.
FWIW, (ql:quickload :hunchentoot) on os x 10.8.2 fails because the pre-installed maven on that system is version 3.0.3: Error loading /Users/rudi/Library/Caches/common-lisp/abcl-1.1.0-dev-fasl39-macosx-java/Users/rudi/Library/quicklisp/dists/quicklisp/software/cffi_0.10.7.1/src/cffi-abcl.abcl at line 8 (offset 577) #<THREAD "interpreter" {67D77351}>: Debugger invoked on condition of type SIMPLE-ERROR We need maven-3.0.4 or later. Restarts: 0: RETRY Retry compiling #<ASDF:MVN "jna" "net.java.dev.jna/jna/3.5.1">. Could we make do with maven 3.0.3? (Upgrading to Oracle's jdk 1.7 causes other stuff to break on my machine, some of it work-related.) Rudi
On 11/26/12 3:46 PM, Rudolf Schlatte wrote:
On Nov 26, 2012, at 15:11, Mark Evenson <evenson@panix.com> wrote:
Overall, I think we are more-or-less ready to release the long awaited abcl-1.1.0 with the exception of resolving the outstanding ANSI regression failures.
Hooray!
[...]
Additionally, we should probably check the Novemeber Quicklisp test reports to see if anything has been broken. I seem to see a problem with Hunchentoot under SLIME regarding the PATHNAME reader, but this may be something local to me.
FWIW, (ql:quickload :hunchentoot) on os x 10.8.2 fails because the pre-installed maven on that system is version 3.0.3:
Error loading /Users/rudi/Library/Caches/common-lisp/abcl-1.1.0-dev-fasl39-macosx-java/Users/rudi/Library/quicklisp/dists/quicklisp/software/cffi_0.10.7.1/src/cffi-abcl.abcl at line 8 (offset 577) #<THREAD "interpreter" {67D77351}>: Debugger invoked on condition of type SIMPLE-ERROR We need maven-3.0.4 or later. Restarts: 0: RETRY Retry compiling #<ASDF:MVN "jna" "net.java.dev.jna/jna/3.5.1">.
Could we make do with maven 3.0.3? (Upgrading to Oracle's jdk 1.7 causes other stuff to break on my machine, some of it work-related.)
Rudi
Agreed that abcl "just working" on current OS X is a good thing to target. quid pro quo: I'll take a look at [#268][] if you check out [#270][] for fifteen minutes? Partly, I'd like your take on how the current ABCL-TEST-LISP suite works for you. I suspect that some documentation in this direction based on questions for you would be a good idea… [#268]: http://trac.common-lisp.net/armedbear/ticket/268 [#270]: http://trac.common-lisp.net/armedbear/ticket/270 -- "A screaming comes across the sky. It has happened before, but there is nothing to compare it to now."
On Nov 26, 2012, at 18:05, Mark Evenson <evenson@panix.com> wrote:
quid pro quo: I'll take a look at [#268][] if you check out [#270][] for fifteen minutes? Partly, I'd like your take on how the current ABCL-TEST-LISP suite works for you. I suspect that some documentation in this direction based on questions for you would be a good idea…
[#268]: http://trac.common-lisp.net/armedbear/ticket/268 [#270]: http://trac.common-lisp.net/armedbear/ticket/270
I fixed the bug -- Java-side %documentation didn't know about generic functions. I didn't know much about abcl's test suite itself, since I mostly just run ansi-tests manually, so I added a sentence to README to help me remember it. (Maybe we should rename README to README.md? It seems to be valid markdown.) Rudi
On 11/26/12 8:53 PM, Rudolf Schlatte wrote:
I fixed the bug -- Java-side %documentation didn't know about generic functions. I didn't know much about abcl's test suite itself, since I mostly just run ansi-tests manually, so I added a sentence to README to help me remember it. (Maybe we should rename README to README.md? It seems to be valid markdown.)
Unfortunately your commit to fix the DOCUMENTATION bug [has broken tests concerning DESCRIBE in ANSI][#272]. Could you please take a look when you get a chance? [#272]: http://trac.common-lisp.net/armedbear/ticket/272 With [r14529][], maven-3.0.3 should work again. [r14529]: http://trac.common-lisp.net/armedbear/changeset/14259 -- "A screaming comes across the sky. It has happened before, but there is nothing to compare to it now."
On Nov 28, 2012, at 15:43 , Mark Evenson wrote:
Unfortunately your commit to fix the DOCUMENTATION bug [has broken tests concerning DESCRIBE in ANSI][#272]. Could you please take a look when you get a chance?
Sorry, will do.
With [r14529][], maven-3.0.3 should work again.
[r14529]: http://trac.common-lisp.net/armedbear/changeset/14259
Many thanks! Rudi
On Nov 28, 2012, at 15:43, Mark Evenson <evenson@panix.com> wrote:
Unfortunately your commit to fix the DOCUMENTATION bug [has broken tests concerning DESCRIBE in ANSI][#272]. Could you please take a look when you get a chance?
This should be fixed now. We're back to 12 ansi-test failures. Rudi
Dear ABCL developers, I apologize if this is not the right channel for this kind of request, but I'd like to suggest the following change to src/org/armedbear/lisp/Lisp.java (for CFFI to identify the platform properly on MacOSX). @@ -2378,9 +2378,9 @@ } // Processor architecture if(osArch != null) { - if (osArch.equals("amd64")) + if (osArch.equals("amd64")||osArch.equals("x86_64")) featureList = new Cons(Keyword.X86_64, featureList); - else if (osArch.equals("x86")) + else if (osArch.equals("x86")||osArch.equals("i386")) featureList = new Cons(Keyword.X86, featureList); } Symbol.FEATURES.initializeSpecial(featureList); Also, there is a problem with contrib/jfli/test/yanking.lisp, there is a referece to abclidea:*lisp-dir* and that package is not part of ABCL. As a workaround I've changed contrib/jfli/jfli.asd to remove the (:module test :components ((:file "yanking")). Best regards, Carlos On Mon, Nov 26, 2012 at 3:11 PM, Mark Evenson <evenson@panix.com> wrote:
Overall, I think we are more-or-less ready to release the long awaited abcl-1.1.0 with the exception of resolving the outstanding ANSI regression failures.
I have moved all previously outstanding issues to the abcl-1.1.1 milestone, and I have added as three blocker tickets to the abcl-1.1.0 milestone to decide on what to do.
http://trac.common-lisp.net/armedbear/ticket/265 http://trac.common-lisp.net/armedbear/ticket/266 http://trac.common-lisp.net/armedbear/ticket/267
For each of these issues, I have identified the changeset that caused the regression via bisecting the source tree. If the result of the bisection is correct, all three issues should be relatively easy to track down, as "not much happened" in the commits. But maybe something was screwy in my process (I did it "on the road" from an iPad ssh client last week in a non-traditional--for me--working environment).
Additionally, we should probably check the Novemeber Quicklisp test reports to see if anything has been broken. I seem to see a problem with Hunchentoot under SLIME regarding the PATHNAME reader, but this may be something local to me.
I need to finish writing up the CHANGES, and make a quick pass through the manual, updating the grovelled docstring sections.
Hopefully, we can have a quick consultation on #abcl tonight at 2100 CST to see if these plans make sense.
-- "A screaming comes across the sky. It has happened before but there is nothing to compare to it now."
_______________________________________________ armedbear-devel mailing list armedbear-devel@common-lisp.net http://lists.common-lisp.net/cgi-bin/mailman/listinfo/armedbear-devel
On Nov 26, 2012, at 11:34 PM, Carlos Ungil <carlos.ungil@gmail.com> wrote:
Dear ABCL developers,
I apologize if this is not the right channel for this kind of request, but I'd like to suggest the following change to src/org/armedbear/lisp/Lisp.java (for CFFI to identify the platform properly on MacOSX).
@@ -2378,9 +2378,9 @@ } // Processor architecture if(osArch != null) { - if (osArch.equals("amd64")) + if (osArch.equals("amd64")||osArch.equals("x86_64")) featureList = new Cons(Keyword.X86_64, featureList); - else if (osArch.equals("x86")) + else if (osArch.equals("x86")||osArch.equals("i386")) featureList = new Cons(Keyword.X86, featureList); } Symbol.FEATURES.initializeSpecial(featureList);
Erik applied this in [r14256][]. Thanks for the patch! [r14256]: http://trac.common-lisp.net/armedbear/changeset/14256
Also, there is a problem with contrib/jfli/test/yanking.lisp, there is a referece to abclidea:*lisp-dir* and that package is not part of ABCL. As a workaround I've changed contrib/jfli/jfli.asd to remove the (:module test :components ((:file "yanking"))
Since there is no need for runtime dependency for JFLI on IntelliJ, I [have moved the JFLI IntelliJ test suite into its own ASDF system definition][r14257]. Please let us know if this does' work for you for some reason. [r14257]: http://trac.common-lisp.net/armedbear/changeset/14257 -- "A screaming comes across the sky. It has happened before but there is nothing to compare to it now."
Mark, Erik, thank you for those fixes. There is another bug in jfli.lisp, at line 366 there is a call to sys::ensure-class but it seems that this function has been moved to another package: mop:ensure-class I have two more comments about jfli.lisp (but these are not critical issues). 1) the function unbox-string (never used, probably it can be removed) calls the undefined function convert-from-java-string 2) I used to have an additional clause in infer-box-type (I don't remember the details, but I think I had to add it to be able to pass a Float argument to some java API): ((typep x '(integer -9223372036854775808 +9223372036854775807)) long.type) + ((typep x 'single-float) float.type) ((numberp x) double.type) Cheers, Carlos On Wed, Nov 28, 2012 at 10:26 AM, Mark Evenson <evenson@panix.com> wrote:
On Nov 26, 2012, at 11:34 PM, Carlos Ungil <carlos.ungil@gmail.com> wrote:
Dear ABCL developers,
I apologize if this is not the right channel for this kind of request, but I'd like to suggest the following change to src/org/armedbear/lisp/Lisp.java (for CFFI to identify the platform properly on MacOSX).
@@ -2378,9 +2378,9 @@ } // Processor architecture if(osArch != null) { - if (osArch.equals("amd64")) + if (osArch.equals("amd64")||osArch.equals("x86_64")) featureList = new Cons(Keyword.X86_64, featureList); - else if (osArch.equals("x86")) + else if (osArch.equals("x86")||osArch.equals("i386")) featureList = new Cons(Keyword.X86, featureList); } Symbol.FEATURES.initializeSpecial(featureList);
Erik applied this in [r14256][]. Thanks for the patch!
[r14256]: http://trac.common-lisp.net/armedbear/changeset/14256
Also, there is a problem with contrib/jfli/test/yanking.lisp, there is a referece to abclidea:*lisp-dir* and that package is not part of ABCL. As a workaround I've changed contrib/jfli/jfli.asd to remove the (:module test :components ((:file "yanking"))
Since there is no need for runtime dependency for JFLI on IntelliJ, I [have moved the JFLI IntelliJ test suite into its own ASDF system definition][r14257]. Please let us know if this does' work for you for some reason.
[r14257]: http://trac.common-lisp.net/armedbear/changeset/14257
-- "A screaming comes across the sky. It has happened before but there is nothing to compare to it now."
On 11/28/12 11:10 PM, Carlos Ungil wrote:
Mark, Erik,
thank you for those fixes.
There is another bug in jfli.lisp, at line 366 there is a call to sys::ensure-class but it seems that this function has been moved to another package: mop:ensure-class
Addressed in r14295. [r14295]: http://trac.common-lisp.net/armedbear/changeset/14295
I have two more comments about jfli.lisp (but these are not critical issues).
1) the function unbox-string (never used, probably it can be removed) calls the undefined function convert-from-java-string
2) I used to have an additional clause in infer-box-type (I don't remember the details, but I think I had to add it to be able to pass a Float argument to some java API):
((typep x '(integer -9223372036854775808 +9223372036854775807)) long.type) + ((typep x 'single-float) float.type) ((numberp x) double.type)
The ABCL version of JFLI could use some patches by someone who uses it, a quality I do no currently posses. If patches are submitted, I will review to patch, but otherwise I unfortunately don't have a lot of time to spend here. -- "A screaming comes across the sky. It has happened before, but there is nothing to compare it to now."
I will run cl-test-grid tests on the latest ABCL revision on the last quicklisp to compare it with the previous ABCL version and ensure there is no regressions. It will take several days or maybe a week, because right now I am collecting test results from all other lisps (including previous ABCL revisions). Best regards, - Anton
Hello. I've just updated http://common-lisp.net/project/cl-test-grid/abcl.html with comparision between the previous release and version abcl-1.2.0-dev-svn-14270-fasl39-linux-x86. As you can see, the only regresssion is the timeout reported as ticket #274. Best regards, - Anton
03.12.2012, 23:05, "Anton Vodonosov" <avodonosov@yandex.ru>:
Hello.
I've just updated http://common-lisp.net/project/cl-test-grid/abcl.html with comparision between the previous release and version abcl-1.2.0-dev-svn-14270-fasl39-linux-x86.
As you can see, the only regresssion is the timeout reported as ticket #274.
Best regards, - Anton
BTW, I run tests without CFFI now, because the machine doesn't have maven installed to make JNA available. One of these days I want to solve this and make JNA available, and rerun the ABCL tests now with CFFI.
On 12/3/12 8:13 PM, Anton Vodonosov wrote:
03.12.2012, 23:05, "Anton Vodonosov" <avodonosov@yandex.ru>:
[…]
BTW, I run tests without CFFI now, because the machine doesn't have maven installed to make JNA available.
One of these days I want to solve this and make JNA available, and rerun the ABCL tests now with CFFI.
CFFI needs a release to Quicklisp: changes already accepted by Luis, et. al. Macports ABCL now has a "+ffi" variant that will ensure that the proper Apache Maven binaries are installed. -- "A screaming comes across the sky. It has happened before, but there is nothing to compare it to now."
On 12/3/12 8:05 PM, Anton Vodonosov wrote:
Hello.
I've just updated http://common-lisp.net/project/cl-test-grid/abcl.html with comparision between the previous release and version abcl-1.2.0-dev-svn-14270-fasl39-linux-x86.
As you can see, the only regresssion is the timeout reported as ticket #274.
Thanks! With these results, I have confidence in the current abcl-1.1.0 release candidates. Go for release on my end. [abcl-1.1.0]: http://trac.common-lisp.net/armedbear/milestone/1.1.0 -- "A screaming comes across the sky. It has happened before, but there is nothing to compare it to now."
On Dec 3, 2012, at 20:24, Mark Evenson <evenson@panix.com> wrote:
On 12/3/12 8:05 PM, Anton Vodonosov wrote:
Hello.
I've just updated http://common-lisp.net/project/cl-test-grid/abcl.html with comparision between the previous release and version abcl-1.2.0-dev-svn-14270-fasl39-linux-x86.
As you can see, the only regresssion is the timeout reported as ticket #274.
Thanks! With these results, I have confidence in the current abcl-1.1.0 release candidates.
Go for release on my end.
Can I have 1-2 days to check the two MOP bugs Pascal reported? It would be neat to have them fixed, if the necessary changes don't touch anything outside CLOS. Rudi
On 12/3/12 8:26 PM, Rudolf Schlatte wrote:
On Dec 3, 2012, at 20:24, Mark Evenson <evenson@panix.com> wrote:
On 12/3/12 8:05 PM, Anton Vodonosov wrote:
Hello.
I've just updated http://common-lisp.net/project/cl-test-grid/abcl.html with comparision between the previous release and version abcl-1.2.0-dev-svn-14270-fasl39-linux-x86.
As you can see, the only regresssion is the timeout reported as ticket #274.
Thanks! With these results, I have confidence in the current abcl-1.1.0 release candidates.
Go for release on my end.
Can I have 1-2 days to check the two MOP bugs Pascal reported? It would be neat to have them fixed, if the necessary changes don't touch anything outside CLOS.
Sure. But we've got abcl-1.1.1 30 days later if you find that you need more time? RC3 at the end of this week? -- "A screaming comes across the sky. It has happened before, but there is nothing to compare it to now."
On Dec 3, 2012, at 20:29, Mark Evenson <evenson@panix.com> wrote:
On 12/3/12 8:26 PM, Rudolf Schlatte wrote:
On Dec 3, 2012, at 20:24, Mark Evenson <evenson@panix.com> wrote:
On 12/3/12 8:05 PM, Anton Vodonosov wrote:
Hello.
I've just updated http://common-lisp.net/project/cl-test-grid/abcl.html with comparision between the previous release and version abcl-1.2.0-dev-svn-14270-fasl39-linux-x86.
As you can see, the only regresssion is the timeout reported as ticket #274.
Thanks! With these results, I have confidence in the current abcl-1.1.0 release candidates.
Go for release on my end.
Can I have 1-2 days to check the two MOP bugs Pascal reported? It would be neat to have them fixed, if the necessary changes don't touch anything outside CLOS.
Sure. But we've got abcl-1.1.1 30 days later if you find that you need more time? RC3 at the end of this week?
Sounds good. Rudi
On Dec 3, 2012, at 20:30, Rudolf Schlatte <rudi@constantly.at> wrote:
On Dec 3, 2012, at 20:29, Mark Evenson <evenson@panix.com> wrote:
On 12/3/12 8:26 PM, Rudolf Schlatte wrote:
Can I have 1-2 days to check the two MOP bugs Pascal reported? It would be neat to have them fixed, if the necessary changes don't touch anything outside CLOS.
Sure. But we've got abcl-1.1.1 30 days later if you find that you need more time? RC3 at the end of this week?
I've checked in fixes for the two reported bugs now. Rudi
On 12/5/12 4:19 PM, Rudolf Schlatte wrote: […
I've checked in fixes for the two reported bugs now.
Rudi
Both changesets have been backported to what will be abcl-1.1.0-rc-2. Thanks for the fixes. http://trac.common-lisp.net/armedbear/changeset/14290 http://trac.common-lisp.net/armedbear/changeset/14294 -- "A screaming comes across the sky. It has happened before, but there is nothing to compare it to now."
03.12.2012, 23:21, "Anton Vodonosov" <avodonosov@yandex.ru>:
Hello.
I've just updated http://common-lisp.net/project/cl-test-grid/abcl.html with comparision between the previous release and version abcl-1.2.0-dev-svn-14270-fasl39-linux-x86.
As you can see, the only regresssion is the timeout reported as ticket #274.
I have updated the page with results of abcl-1.2.0-dev-svn-14300-fasl39-linux-x86. No new regressions .
participants (5)
-
Anton Vodonosov -
Carlos Ungil -
Mark Evenson -
Rudi Schlatte -
Rudolf Schlatte