Bug#402736: FTBFS on alpha on both Etch and Sid
retitle 402736 FTBFS on alpha: memory allocation on 64-bit architectures tags 402736 pending thanks On Tue, Dec 12, 2006 at 09:47:06PM +0100, Aurélien GÉRÔME wrote: Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 16384 (LWP 31355)] 0x000120085700 in cresmatvar_ (id=0x11fec55c0, lw=0x11fec558c, str=0x120fbbe9e, lstr=0x11fec5628, __g77_length_str=4096) at inisci.f:427 427 istk(il ) = 10 (gdb) bt #0 0x000120085700 in cresmatvar_ (id=0x11fec55c0, lw=0x11fec558c, str=0x120fbbe9e, lstr=0x11fec5628, __g77_length_str=4096) at inisci.f:427 #1 0x000120084780 in inisci_ (ini1=0x120aaf3b8, vsizr=0x11fec5690, ierr=0x120ab1c5c) at inisci.f:259 #2 0x0001207812c0 in VTRun () #3 0x000120786794 in main_sci () #4 0x000120787018 in realmain_ () #5 0x00012004d720 in MAIN__ () at mainsci.f:4 I should have looked at the Debian patches earlier, because I found the fix. :) On 64-bit architectures, specific memory allocation functions are used in scilab. It is already the case on ia64 and amd64, so I added alpha to that list. This is probably because of some issue with the mix of Fortran and C, though I have not looked in deeper... Okay, so let's go for a rebuild of all architectures and a hint to get scilab into Etch! ;) Cheers, -- .''`. Aurélien GÉRÔME : :' : `. `'` Free Software Developer `- Unix Sys Net Admin signature.asc Description: Digital signature
Bug#402736: FTBFS on alpha on both Etch and Sid
Le mardi 12 décembre 2006 à 12:29 +0100, Aurélien GÉRÔME a écrit : Package: scilab Version: 4.0-11 Severity: serious Unfortunately, scilab FTBFS on alpha on both Etch and Sid with differents errors. :( Firstly, Sid: * A first FTBFS occurs in the file ./routines/Javasci/classes/BadDataArgumentException.java interpreted by javac (gcj-wrapper-4.1). This is easily fixed by replacing the accentuated character in the French comment (s/gérer/gerer/). Thanks for this bug report. I translated the comment into english in the upstream but this change won't be taken in account in the version 4.1 of Scilab (which should be released today or tomorrow). I also made this change in the trunk of Scilab. I commited the change also in the debian svn for scilab but I saw here http://packages.qa.debian.org/s/scilab.html Not touching package, as requested by freeze (contact debian-release if update is needed) What should I do ? * A second FTBFS occurs afterwards in the file ./routines/Javasci/javasci_SciStringArray.c, see the attached build log. Java is not mandatory in Scilab. It is used when you need to use the scilab engine from a Java application (this could become a new debian package btw). Scilab compilation complains about the jni.h file. Maybe if you install libgcj4-dev, this may work but I am not sure about that. Secondly, Etch: * A FTBFS occurs via the oom-killer during the use of the freshly built ./bin/scilab binary to generate HTML documentation, see the attached build log. I saw that bug before : http://www.scilab.org/cgi-bin/bugzilla_bug_II/show_bug.cgi?id=1483 I forwarded this error to the Scilab Mailing list. Cheers, Sylvestre
Bug#402736: FTBFS on alpha on both Etch and Sid
On Tue, Dec 12, 2006 at 02:46:59PM +0100, Sylvestre Ledru wrote: Le mardi 12 décembre 2006 à 12:29 +0100, Aurélien GÉRÔME a écrit : * A first FTBFS occurs in the file ./routines/Javasci/classes/BadDataArgumentException.java interpreted by javac (gcj-wrapper-4.1). This is easily fixed by replacing the accentuated character in the French comment (s/gérer/gerer/). Thanks for this bug report. I translated the comment into english in the upstream but this change won't be taken in account in the version 4.1 of Scilab (which should be released today or tomorrow). I also made this change in the trunk of Scilab. I commited the change also in the debian svn for scilab but I saw here http://packages.qa.debian.org/s/scilab.html Not touching package, as requested by freeze (contact debian-release if update is needed) What should I do ? Indeed, this is not a necessary change in the Debian packaging, because we do not provide the Java interface anyway. Moreover, it seems that this is a gcj bug. I had gcj in my manual build environment, the configure script detected it, and used it. That is why I said earlier that we might consider disabling it manually, instead of relying on automatic detection by the configure script. * A second FTBFS occurs afterwards in the file ./routines/Javasci/javasci_SciStringArray.c, see the attached build log. Java is not mandatory in Scilab. It is used when you need to use the scilab engine from a Java application (this could become a new debian package btw). Scilab compilation complains about the jni.h file. Maybe if you install libgcj4-dev, this may work but I am not sure about that. Yes, this is probably due to wrong headers, thought I did not search deeper when I read Torsten's feedback on it, i.e. we do not use it. :) * A FTBFS occurs via the oom-killer during the use of the freshly built ./bin/scilab binary to generate HTML documentation, see the attached build log. I saw that bug before : http://www.scilab.org/cgi-bin/bugzilla_bug_II/show_bug.cgi?id=1483 I forwarded this error to the Scilab Mailing list. Sure, this is now the real bug which happens on both Etch and Sid on alpha architecture. However, it happens at an early stage in scilab (when you launch it), so it probably has something to do with either wrong compiled code for alpha (it would be a gcc bug) or an architecture-specific quirk. I am currently investigating the issue with gdb... Cheers, -- .''`. Aurélien GÉRÔME : :' : `. `'` Free Software Developer `- Unix Sys Net Admin signature.asc Description: Digital signature
Bug#402736: FTBFS on alpha on both Etch and Sid
On Tue, Dec 12, 2006 at 03:10:09PM +0100, Aurélien GÉRÔME wrote: Sure, this is now the real bug which happens on both Etch and Sid on alpha architecture. However, it happens at an early stage in scilab (when you launch it), so it probably has something to do with either wrong compiled code for alpha (it would be a gcc bug) or an architecture-specific quirk. I am currently investigating the issue with gdb... Starting program: /home/ag/build/scilab-4.0/bin/scilex [Thread debugging using libthread_db enabled] [New Thread 16384 (LWP 17888)] [New Thread 32769 (LWP 17893)] [New Thread 16386 (LWP 17894)] Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 16384 (LWP 17888)] 0x000120078cfc in cresmatvar_ () (gdb) bt #0 0x000120078cfc in cresmatvar_ () #1 0x000120078418 in inisci_ () #2 0x0001205771a8 in VTRun () #3 0x00012057c67c in main_sci () #4 0x00012057cf00 in realmain_ () #5 0x00012004d718 in MAIN__ () The bug is located in the Fortran code of ./routines/system/inisci.f, so it must be related to g77 in some way. I searched a bit, but I do not know much about Fortran. Sylvestre, can you be of any help on this? :) Cheers, -- .''`. Aurélien GÉRÔME : :' : `. `'` Free Software Developer `- Unix Sys Net Admin signature.asc Description: Digital signature
Bug#402736: FTBFS on alpha on both Etch and Sid
On Tue, Dec 12, 2006 at 08:10:07PM +0100, Aurélien GÉRÔME wrote: Starting program: /home/ag/build/scilab-4.0/bin/scilex [Thread debugging using libthread_db enabled] [New Thread 16384 (LWP 17888)] [New Thread 32769 (LWP 17893)] [New Thread 16386 (LWP 17894)] Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 16384 (LWP 17888)] 0x000120078cfc in cresmatvar_ () (gdb) bt #0 0x000120078cfc in cresmatvar_ () #1 0x000120078418 in inisci_ () #2 0x0001205771a8 in VTRun () #3 0x00012057c67c in main_sci () #4 0x00012057cf00 in realmain_ () #5 0x00012004d718 in MAIN__ () The bug is located in the Fortran code of ./routines/system/inisci.f, so it must be related to g77 in some way. I searched a bit, but I do not know much about Fortran. Sylvestre, can you be of any help on this? :) I rebuilt scilab with FC_OPTIONS=-g and here is the gdb session with debugging symbols support. (gdb) r Starting program: /home/ag/build/scilab-4.0/bin/scilex [Thread debugging using libthread_db enabled] [New Thread 16384 (LWP 31355)] [New Thread 32769 (LWP 31360)] [New Thread 16386 (LWP 31361)] Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 16384 (LWP 31355)] 0x000120085700 in cresmatvar_ (id=0x11fec55c0, lw=0x11fec558c, str=0x120fbbe9e, lstr=0x11fec5628, __g77_length_str=4096) at inisci.f:427 427 istk(il ) = 10 (gdb) bt #0 0x000120085700 in cresmatvar_ (id=0x11fec55c0, lw=0x11fec558c, str=0x120fbbe9e, lstr=0x11fec5628, __g77_length_str=4096) at inisci.f:427 #1 0x000120084780 in inisci_ (ini1=0x120aaf3b8, vsizr=0x11fec5690, ierr=0x120ab1c5c) at inisci.f:259 #2 0x0001207812c0 in VTRun () #3 0x000120786794 in main_sci () #4 0x000120787018 in realmain_ () #5 0x00012004d720 in MAIN__ () at mainsci.f:4 Cheers, -- .''`. Aurélien GÉRÔME : :' : `. `'` Free Software Developer `- Unix Sys Net Admin signature.asc Description: Digital signature