Re: [Fink-devel] problems with g77 3.4 segmentation faults
Hi Jeff et al: Many thanks again for your help with this. Although this version did in fact get rid of my segmentation fault, it didn't solve my other problem, which means the other problem was not specific to compiler version, so at least I have learned something. Given this, and what Remi Mommsen reports, I cannot make a compelling argument in favor of using version 3.3.2 instead of 3.4. (The other problem occurs in a package that isn't in fink, and I have now found can be solved using IBM's xlf.) I'll submit a bug report to the g77 folks for the seg fault in 3.4. I haven't yet had a chance to see if it is platform-indepenent. Again, thanks for helping with this. Bill Scott On Nov 8, 2003, at 5:12 PM, Remi Mommsen wrote: Hi Jeff, I tried your new g77 3.3.2 on 10.3 with cernlib. cernlib builds fine, but the test fails as it was the case with 3.3(.1) (see our email exchange from Jul 17, 2003). With the g77 3.4-20031015 it works flawlessly. Thanks for your effort. Remi William G. Scott Associate Professor Department of Chemistry and Biochemistry and The Center for the Molecular Biology of RNA Sinsheimer Laboratories University of California at Santa Cruz Santa Cruz, California 95064 USA phone: +1-831-459-5367 (office) +1-831-459-5292 (lab) fax: +1-831-4593139 (fax) url: http://chemistry.ucsc.edu/~wgscott/
Re: [Fink-devel] problems with g77 3.4 segmentation faults
Hi Jeff, I tried your new g77 3.3.2 on 10.3 with cernlib. cernlib builds fine, but the test fails as it was the case with 3.3(.1) (see our email exchange from Jul 17, 2003). With the g77 3.4-20031015 it works flawlessly. Thanks for your effort. Remi On Friday, November 7, 2003, at 04:43 AM, Jeff Whitaker wrote: William and Martin: I've verified that the latest released g77 (3.3.2) does not have the assembler bug present in 3.1,3.3.1 and Apple's 3.3-20030304. As Martin pointed out, the reason I put the 3.4 package in, even though it is not a released version of g77, is that it was the only version that didn't have the bug at the time. William - I've put 3.3.2 in 10.3 unstable, please try it on your packages, if it works for them (and all the other fortran packages) I can make it the primary g77 version either by using Martin's trick, or adding Epoch: 1 to the info file. -Jeff On Fri, 7 Nov 2003, Martin Costabel wrote: William Scott wrote: [] Is there any way to have version 3.1 and 3.3 cohabitating in 10.3? I don't think there is a reasonable way to have two *installed* versions of g77 cohabitate. For two *package descriptions* cohabitating, the right way would be to create a g77-3.1 package, i.e. instead of Package: g77 Version: 3.1-20020420 Revision: 6 make it Package: g77-3.1 Version: 20020420 Revision: 6 and make it conflict/replace with g77. BUT: Whereas the fatal bug of g77-3.1 is clearly documented(*) and has not gone away with Panther, the bugs with g77-3.4 that have been reported here do not seem to be deterministic. At least I haven't seen bug reports that can be reproduced by everyone. I don't think Jeff or anyone else feels the need to have the very latest g77 version. The only criterion is to have a fortran compiler that just works. And so far the best strategy seemed to be to use recent snapshots in order to take advantage of the bug fixes that went into the official gcc sources. I've noticed that some other people have had problems too. Alternatively, is there a way for me to install 3.1 and avoid having fink try to auto-update it to 3.4 every time I issue fink update-all? The latter would solve my problem, but I still can't put some of my packages into 10.3 until I can get them to compile. (*) Contrary to what some have reported here, the gcc3.3 of Panther still has the bug where a trivial program defining an empty struct does not compile. Compiling the 2-liner struct { } foo = { }; void * bar(void) { return foo; } with gcc-3.3 gives costabel% gcc test_c.c /var/tmp//cc7djqBM.s:22:section difference relocatable subtraction expression, _foo minus L001$pb using a symbol at the end of section will not produce an assembly time constant /var/tmp//cc7djqBM.s:22:use a symbol with a constant value created with an assignment instead of the expression, L_const_sym = _foo - L001$pb /var/tmp//cc7djqBM.s:21:section difference relocatable subtraction expression, _foo minus L001$pb using a symbol at the end of section will not produce an assembly time constant /var/tmp//cc7djqBM.s:21:use a symbol with a constant value created with an assignment instead of the expression, L_const_sym = _foo - L001$pb With g77-3.1, the following program gives the same error message. Even a program consisting only of 1 line end gives the same error: costabel% cat test_f.f program hello write(6, '(1X,A)') 'Hello world' end [abook:]costabel% g77 test_f.f /var/tmp//ccAvjI34.s:48:section difference relocatable subtraction expression, LC3 minus L1$pb using a symbol at the end of section will not produce an assembly time constant /var/tmp//ccAvjI34.s:48:use a symbol with a constant value created with an assignment instead of the expression, L_const_sym = LC3 - L1$pb /var/tmp//ccAvjI34.s:47:section difference relocatable subtraction expression, LC3 minus L1$pb using a symbol at the end of section will not produce an assembly time constant /var/tmp//ccAvjI34.s:47:use a symbol with a constant value created with an assignment instead of the expression, L_const_sym = LC3 - L1$pb The accepted way to correct this bug with g77 prior to 3.4 was to replace the assembler in /usr/libexec/gcc/darwin/ppc/ by an older one from the Dec2002 dev tools (or to install an as package that does this), and I don't think anyone wants to do this on Panther. -- Jeffrey S. Whitaker Phone : (303)497-6313 NOAA/OAR/CDC R/CDC1FAX : (303)497-6449 325 BroadwayWeb : http://www.cdc.noaa.gov/~jsw Boulder, CO, USA 80305-3328 Office: Skaggs Research Cntr 1D-124 --- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ ___ Fink-devel mailing list [EMAIL
Re: [Fink-devel] problems with g77 3.4 segmentation faults
William Scott wrote: [] Is there any way to have version 3.1 and 3.3 cohabitating in 10.3? I don't think there is a reasonable way to have two *installed* versions of g77 cohabitate. For two *package descriptions* cohabitating, the right way would be to create a g77-3.1 package, i.e. instead of Package: g77 Version: 3.1-20020420 Revision: 6 make it Package: g77-3.1 Version: 20020420 Revision: 6 and make it conflict/replace with g77. BUT: Whereas the fatal bug of g77-3.1 is clearly documented(*) and has not gone away with Panther, the bugs with g77-3.4 that have been reported here do not seem to be deterministic. At least I haven't seen bug reports that can be reproduced by everyone. I don't think Jeff or anyone else feels the need to have the very latest g77 version. The only criterion is to have a fortran compiler that just works. And so far the best strategy seemed to be to use recent snapshots in order to take advantage of the bug fixes that went into the official gcc sources. I've noticed that some other people have had problems too. Alternatively, is there a way for me to install 3.1 and avoid having fink try to auto-update it to 3.4 every time I issue fink update-all? The latter would solve my problem, but I still can't put some of my packages into 10.3 until I can get them to compile. (*) Contrary to what some have reported here, the gcc3.3 of Panther still has the bug where a trivial program defining an empty struct does not compile. Compiling the 2-liner struct { } foo = { }; void * bar(void) { return foo; } with gcc-3.3 gives costabel% gcc test_c.c /var/tmp//cc7djqBM.s:22:section difference relocatable subtraction expression, _foo minus L001$pb using a symbol at the end of section will not produce an assembly time constant /var/tmp//cc7djqBM.s:22:use a symbol with a constant value created with an assignment instead of the expression, L_const_sym = _foo - L001$pb /var/tmp//cc7djqBM.s:21:section difference relocatable subtraction expression, _foo minus L001$pb using a symbol at the end of section will not produce an assembly time constant /var/tmp//cc7djqBM.s:21:use a symbol with a constant value created with an assignment instead of the expression, L_const_sym = _foo - L001$pb With g77-3.1, the following program gives the same error message. Even a program consisting only of 1 line end gives the same error: costabel% cat test_f.f program hello write(6, '(1X,A)') 'Hello world' end [abook:]costabel% g77 test_f.f /var/tmp//ccAvjI34.s:48:section difference relocatable subtraction expression, LC3 minus L1$pb using a symbol at the end of section will not produce an assembly time constant /var/tmp//ccAvjI34.s:48:use a symbol with a constant value created with an assignment instead of the expression, L_const_sym = LC3 - L1$pb /var/tmp//ccAvjI34.s:47:section difference relocatable subtraction expression, LC3 minus L1$pb using a symbol at the end of section will not produce an assembly time constant /var/tmp//ccAvjI34.s:47:use a symbol with a constant value created with an assignment instead of the expression, L_const_sym = LC3 - L1$pb The accepted way to correct this bug with g77 prior to 3.4 was to replace the assembler in /usr/libexec/gcc/darwin/ppc/ by an older one from the Dec2002 dev tools (or to install an as package that does this), and I don't think anyone wants to do this on Panther. -- Martin --- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ ___ Fink-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/fink-devel
Re: [Fink-devel] problems with g77 3.4 segmentation faults
William and Martin: I've verified that the latest released g77 (3.3.2) does not have the assembler bug present in 3.1,3.3.1 and Apple's 3.3-20030304. As Martin pointed out, the reason I put the 3.4 package in, even though it is not a released version of g77, is that it was the only version that didn't have the bug at the time. William - I've put 3.3.2 in 10.3 unstable, please try it on your packages, if it works for them (and all the other fortran packages) I can make it the primary g77 version either by using Martin's trick, or adding Epoch: 1 to the info file. -Jeff On Fri, 7 Nov 2003, Martin Costabel wrote: William Scott wrote: [] Is there any way to have version 3.1 and 3.3 cohabitating in 10.3? I don't think there is a reasonable way to have two *installed* versions of g77 cohabitate. For two *package descriptions* cohabitating, the right way would be to create a g77-3.1 package, i.e. instead of Package: g77 Version: 3.1-20020420 Revision: 6 make it Package: g77-3.1 Version: 20020420 Revision: 6 and make it conflict/replace with g77. BUT: Whereas the fatal bug of g77-3.1 is clearly documented(*) and has not gone away with Panther, the bugs with g77-3.4 that have been reported here do not seem to be deterministic. At least I haven't seen bug reports that can be reproduced by everyone. I don't think Jeff or anyone else feels the need to have the very latest g77 version. The only criterion is to have a fortran compiler that just works. And so far the best strategy seemed to be to use recent snapshots in order to take advantage of the bug fixes that went into the official gcc sources. I've noticed that some other people have had problems too. Alternatively, is there a way for me to install 3.1 and avoid having fink try to auto-update it to 3.4 every time I issue fink update-all? The latter would solve my problem, but I still can't put some of my packages into 10.3 until I can get them to compile. (*) Contrary to what some have reported here, the gcc3.3 of Panther still has the bug where a trivial program defining an empty struct does not compile. Compiling the 2-liner struct { } foo = { }; void * bar(void) { return foo; } with gcc-3.3 gives costabel% gcc test_c.c /var/tmp//cc7djqBM.s:22:section difference relocatable subtraction expression, _foo minus L001$pb using a symbol at the end of section will not produce an assembly time constant /var/tmp//cc7djqBM.s:22:use a symbol with a constant value created with an assignment instead of the expression, L_const_sym = _foo - L001$pb /var/tmp//cc7djqBM.s:21:section difference relocatable subtraction expression, _foo minus L001$pb using a symbol at the end of section will not produce an assembly time constant /var/tmp//cc7djqBM.s:21:use a symbol with a constant value created with an assignment instead of the expression, L_const_sym = _foo - L001$pb With g77-3.1, the following program gives the same error message. Even a program consisting only of 1 line end gives the same error: costabel% cat test_f.f program hello write(6, '(1X,A)') 'Hello world' end [abook:]costabel% g77 test_f.f /var/tmp//ccAvjI34.s:48:section difference relocatable subtraction expression, LC3 minus L1$pb using a symbol at the end of section will not produce an assembly time constant /var/tmp//ccAvjI34.s:48:use a symbol with a constant value created with an assignment instead of the expression, L_const_sym = LC3 - L1$pb /var/tmp//ccAvjI34.s:47:section difference relocatable subtraction expression, LC3 minus L1$pb using a symbol at the end of section will not produce an assembly time constant /var/tmp//ccAvjI34.s:47:use a symbol with a constant value created with an assignment instead of the expression, L_const_sym = LC3 - L1$pb The accepted way to correct this bug with g77 prior to 3.4 was to replace the assembler in /usr/libexec/gcc/darwin/ppc/ by an older one from the Dec2002 dev tools (or to install an as package that does this), and I don't think anyone wants to do this on Panther. -- Jeffrey S. Whitaker Phone : (303)497-6313 NOAA/OAR/CDC R/CDC1FAX : (303)497-6449 325 BroadwayWeb : http://www.cdc.noaa.gov/~jsw Boulder, CO, USA 80305-3328 Office: Skaggs Research Cntr 1D-124 --- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ ___ Fink-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/fink-devel