Created attachment 593584 [details] attached patch for gprbuild.spec Description of problem: gprbuild creates configurations using rpath for the linker. Additionally, it is broken for non-English locales, since it parses "gcc -v" output and not "gcc -dumpversion". Version-Release number of selected component (if applicable): 2011-3 How reproducible: rpath: look into generated cgpr locale: LC_ALL=de_DE.UTF-8 gprversion Steps to Reproduce: 1. run gprbuild on some project 2. grep rpath obj/auto.cgpr 3. run LC_ALL=de_DE.UTF-8 gprversion Actual results: step 2: for Run_Path_Option use ("-Wl,-z,origin,-rpath,"); step 3: Invalid setup of the gprconfig knowledge base Exit code: 4 Expected results: no rpath, working with different locales Additional info:
Created attachment 593585 [details] remove rpath
Created attachment 593586 [details] fix version parsing for non-English locales
Thank you for your bug report. Probably I've lost gprbuild-gccversion.patch while updating to 2011. Besides, I've sent this patch to AdaCore already.
gprbuild-2011-4.fc17 has been submitted as an update for Fedora 17. https://admin.fedoraproject.org/updates/gprbuild-2011-4.fc17
(In reply to comment #1) > Created attachment 593585 [details] > remove rpath What does this patch do? Does it completely disable GPRbuild's ability to embed runpaths in binaries, even when it's not a Fedora package being built? Does it do a better job than the -R option?
I prefer use -R or chrpath. So I'm going to testing gprbuild-2011-4 with other packages and accept or delete patch.
Package gprbuild-2011-4.fc17: * should fix your issue, * was pushed to the Fedora 17 testing repository, * should be available at your local mirror within two days. Update it with: # su -c 'yum update --enablerepo=updates-testing gprbuild-2011-4.fc17' as soon as you are able to. Please go to the following url: https://admin.fedoraproject.org/updates/FEDORA-2012-9890/gprbuild-2011-4.fc17 then log in and leave karma (feedback).
gprbuild-2011-4.fc17 has been pushed to the Fedora 17 stable repository. If problems still persist, please make note of it in this bug report.
The problem with getting the GCC version still exists in GPRbuild 2012 and 2013. In at least the locales fi_FI.UTF-8 and de_DE.UTF-8 it fails to find GCC and claims that "libraries are not supported on this platform". Please restore the gccversion patch.
Created attachment 832212 [details] Git patch for the f19 branch
Created attachment 832213 [details] Git patch for the master branch
Created attachment 832214 [details] Git patch for the f19 branch
Created attachment 832215 [details] Git patch for the master branch I mixed up the patches. Now they're labelled correctly. Sorry.
Bjorn, Thank you for the patch. for me it looks like a bug in gcc localization, not in gprbuild. Unfortunately I'm not familiar with German and Finish languages but gcc-Version and gcc-versio doesn't look as translation of "gcc version"... The original regexp filters all non-gcc compilers out which "could" use gcc alias for compatibility for example but -dumpversion doesn't contain any compiler name string at all. Are some raises possible here? Is something like "^gcc[^\s]\S+ (\S+)" could be less dangerous here?
German is famous for its compound words but it can't match Finnish. In English "gcc version 4.8.2" can be read either as "This is GCC. Specifically, it's version 4.8.2." or as "The GCC version is 4.8.2.". In the latter case "gcc version" is a compound and the only problem with the German translation is that it doesn't spell "GCC" with capitals. The Swedish translator seems to have chosen the first interpretation. Otherwise I would have had this problem too. Vietnamese (vi_VN) appears to express "version" in two words: "gcc phiên bản 4.8.2 20131017 (Red Hat 4.8.2-1) (GCC)" The current pattern will select "bản" as the version. Parsing text that's intended for human consumption is inherently unreliable, and trying to do it with localized text is just crazy. If gcc -v is a stable API, then the documentation should tell how its output can be parsed reliably. Otherwise, if -dumpversion isn't acceptable, then the only way this can become anywhere near reliable is to have GPRbuild set the locale to "C" in the subprocess.