Created attachment 319648 [details] test.f90 Description of problem: Date problem reported: 10-6-08 Problem: gcc-4.3.2 / gfortran-4.3.2 (in the forthcoming tech-preview for RHEL5.3) appears to also have the "DW_TAG_common_block has incorrect placement/scope" issues listed in the following gcc upstream bug: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37738 What gfortran is doing currently is: (1) Merges all common blocks into a single one. (2) Places this single common block into the first subroutine where it was used. Jan Kratochvil in BZ 459376, comment #12 noted that these issues are making it so that gcc-gfortran is incompatible with some fixes made to gdb-6.8 (see BZ 459672), which is what we'll be re-basing gdb on when RHEL5.3 ships (Jan is also the reporter of the upstream bug). What is expected from SEG/Engineering: Need to send this up so we can track possible fixes to gcc-4.3. We need to try and squeeze this one in for RHEL5.3 as well. Issue escalated to Support Engineering Group by: kbaxley. kbaxley assigned to issue for LLNL (HPC). Category set to: Devel Tools::GCC In BZ 459376 (see comment #12): engineering noted that there was a gcc bug in gcc-4.3.2 that's making gfortran incompatible with gdb fixes that recently went in. The BZ for the gdb issue is 459762. See comment #8 in that BZ. Version-Release number of selected component (if applicable): How reproducible: always Steps to Reproduce: I have attached the two testcases that were used in the gcc-bugzilla, and I can reproduce this on an x86-64 system running the gfortran-4.3 tech preview packages: gcc43-gfortran-4.3.2-6.el5 Compile the test programs using something like: gfortran43 -g -o testing2 test2.f90 gfortran43 -g -o testing test.f90 Then run readelf -w against each output file. Actual results: Expected results: Additional info: see attached
Created attachment 319649 [details] test2.f90
I agree this is fatal enough to need a rebuild, but I'm still waiting for upstream ack on the patch.
This has been in rawhide for a few days, without problems and got even through gdb testing, so I'd like to respin gcc43 now for this.
(In reply to comment #0) > Jan Kratochvil in BZ 459376, comment #12 noted that these issues are making it > so that gcc-gfortran is incompatible with some fixes made to gdb-6.8 (see BZ > 459672), which is what we'll be re-basing gdb on when RHEL5.3 ships (Jan is > also the reporter of the upstream bug). This support for GDB exists as a patch now but it is not present in any GDB release (upstream/rhel/fedora/archer). It is a GDB fix/feature planned for RHEL-5.4. Still sure the GCC fix makes sense for other non-GDB debuggers. I can also respin GDB for RHEL-5.3 with the common blocks fix from Bug 459762 but still not all of the other Fortran fixes are done+stable in GDB (all the GCC Fortran fixes look as done already for 5.3, though).
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on therefore solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHEA-2009-0077.html