I have uploaded rebuilt packages of the "fftw" source. Advisory suggestion =================== The published version of the armv5tl package "lib64fftwmpi3-3.3.6-1.mga6" includes a broken dependency, to openmpi instead of openmpi2. The other architectures are not affected. The version bump and rebuild of "fftw" appears to have fixed the issue. No other changes were made to the package. Packages ======== fftw-wisdom-3.3.6-2.mga6 lib64fftw-devel-3.3.6-2.mga6 lib64fftw-static-devel-3.3.6-2.mga6 lib64fftw3-3.3.6-2.mga6 lib64fftwmpi3-3.3.6-2.mga6 lib64fftwomp3-3.3.6-2.mga6 Source ====== fftw-3.3.6-2.mga6.src.rpm
Created attachment 9816 [details] Ignore this, wrong log
Corrected advisory suggestion ============================= The published version of the armv5tl package "libfftwmpi3-3.3.6-1.mga6" includes a broken dependency, to openmpi instead of openmpi2. The other architectures are not affected. The version bump and rebuild of "fftw" appears to have fixed the issue. No other changes were made to the package. Packages ======== fftw-wisdom-3.3.6-2.mga6 lib(64)fftw-devel-3.3.6-2.mga6 lib(64)fftw-static-devel-3.3.6-2.mga6 lib(64)fftw3-3.3.6-2.mga6 lib(64)fftwmpi3-3.3.6-2.mga6 lib(64)fftwomp3-3.3.6-2.mga6 Source: fftw-3.3.6-2.mga6.src.rpm
Created attachment 9817 [details] Correct build log showing the problem The dependency problem surfaced during the build of "synfig" which indirectly depends on libfftmpi.
Comment on attachment 9816 [details] Ignore this, wrong log Sorry, wrong file - see next attachment.
Attachment 9816 description: Build log of armv5tl version => Ignore this, wrong log
MGA6-32 on Dell Latitude D600 MATE No installation issues. My knowledge of fourier analysis is non-existant, but at least the following works OK: $ fftw-wisdom -h Usage: fftw-wisdom [options] [sizes] Create wisdom (pre-planned/optimized transforms) for specified sizes, writing wisdom to stdout (or to a file, using -o). Options: -h, --help: print this help -V, --version: print version/copyright info etc .... and $ fftw-wisdom -V fftw-wisdom tool for FFTW version 3.3.6-pl2. Copyright (c) 2003, 2007-14 Matteo Frigo Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology This program is free software; you can redistribute it and/or modify etc.... Unless anyone objects, its OK for me.
CC: (none) => herman.viaeneWhiteboard: (none) => MGA6-32-OK
For these kinds of fixes that only affect ARM, I don't think new Intel builds and QA need to be involved. The sysadmins should be able to fix these directly.
CC: (none) => pterjan
Probing M6/64 David's comment above is right, I think: this should not have become an update for 1586/x64. Looking at it empirically, the issued version: $ urpmq -i lib64fftwmpi3 Name : lib64fftwmpi3 Version : 3.3.6 Release : 1.mga6 $ urpmq -d lib64fftwmpi3 | grep openmpi lib64openmpi20 [note 20, not 2] The update version: Name : lib64fftwmpi3 Version : 3.3.6 Release : 2.mga6 $ urpmq -d lib64fftwmpi3 | grep openmpi lib64openmpi20 [unchanged] Perhaps this is not even relevant. I installed directly from updates_testing fftw-wisdom-3.3.6-2.mga6.x86_64 which did *not* pull in any of the libraries. Installing those explicitly: # urpmi lib64fftw3 lib64fftwmpi3 lib64fftwomp3 *did* pull in also (among other things) lib64openmpi20 2.1.1 4.mga6 x86_64 if this is what we are talking about. BTAIM OKing & validating to move it on.
Keywords: (none) => advisory, validated_updateWhiteboard: MGA6-32-OK => MGA6-32-OK MGA6-64-OKCC: (none) => lewyssmith, sysadmin-bugs
An update for this issue has been pushed to the Mageia Updates repository. https://advisories.mageia.org/MGAA-2017-0126.html
Resolution: (none) => FIXEDStatus: NEW => RESOLVED