Based on an internal mirror of the various repositories (i.e. it might be slightly out of date), I wrote a script to compare the SRPMs (diff'ing the contents) generated during stage4 with SRPMs from other sources (mostly, F15 official ones, but also from SRPMs still in the builder queues). Color coded with diffs and stats:
http://djdelorie.fedorapeople.org/armv7-srpms.html
(uses javascript to show/hide diffs)
Summary:
99 SRPMs are of the .0.armN variety (only 5 I couldn't find the upstream to diff against)
3 SRPMs in stage 4 have the same NVR as GA/updates BUT DIFFER IN CONTENT (glibc and gdesklets vs GA, crash has changelog diffs vs development)
5 SRPMs match rawhide
11 SRPMs show up in the builder queues ("others") but not upstream
0 SRPMs that show up in the builder queues differ in content
30 SRPMs I couldn't find outside stage4 (might be from an older rawhide)
4704 SRPMs match GA or updates (name and content match)
Note: the HTML file is about 1 MB
Many thanks! Very useful report. Actually had on my todo for some time to do something similar to audit stage4 and very glad you have done it already.
ons 2011-09-28 klockan 23:18 -0400 skrev DJ Delorie:
Summary:
99 SRPMs are of the .0.armN variety (only 5 I couldn't find the upstream to diff against)
Most of these are pending for being replaced by updated mainline F15 builds as they become available. But not all. Some coordination is needed here to make sure all gets pushed to mainline.
3 SRPMs in stage 4 have the same NVR as GA/updates BUT DIFFER IN CONTENT (glibc and gdesklets vs GA, crash has changelog diffs vs development)
glibc appears to be an arm hacked one where no tag was added in Release. Needs someone to take the changes and get them mainline.
gdesklets seems very odd that the .tar.gz differ. needs to be investigated what happened and probably rebuilt to correct the unexpected difference.
crash is the most recent mainline F15 build which adds arm support.
5 SRPMs match rawhide
All except ppl looks like F15 builds not yet in updates.
11 SRPMs show up in the builder queues ("others") but not upstream
Most seem to be F15 builds not yet in updates, but not all.
0 SRPMs that show up in the builder queues differ in content
?
is this matches between SRPMS* and stage4/ .src.rpm? Would be quite worried if there was differences between those.
30 SRPMs I couldn't find outside stage4 (might be from an older rawhide)
In those there is two main categories. Which one can be identified by the name of the stage4 folder.
a) Packages thad did not have an f15 build and got their dist tag modified by the build process. Identified by the folder name having a different dist tag than the srpm.
To prune the list from false indications please do the following
* Use the name of the folder where the srpm is instead of the srpm. This to prune out dist tag differences.
* Ignore the dist tag in the package name to further reduce false mismatches in srpm names.
* For each remaining build in the list without an .armN tag, try fetching the build from koji.
* remove the "other" category. it's not really interesting to compare stage4 to what is in the different SRPM folders on arm-temp. Those exists mainly to aid the build queue management and says very little about the origin of the srpm.
The order in which to look for a matching package is
1. F15 GA (OK) 2. F15 updates (OK) 3. F15 koji (OK with remark) 4. F16 / rawhide koji. (needs to be merged to F15)
I do not think we should look for matches elsewhere, at least not for categorization purposes.
and only look at current stage4 package versions. Some packages in stage4 have been replaced by newer builds. The easiest way to get the right list of stage4 srpms is probably "repoquery -i stage4 -a --qf %{sourcerpm} | sort -u"
Regards Henrik
Nice chart! I was poking through the diffs, and in a number of spots It shows:
-ExclusiveArch: %{ix86} x86_64 ppc alpha sparcv9 +ExclusiveArch: %{ix86} x86_64 ppc alpha sparcv9 armv7hl
I thought we were adding armv5tel as well when we added armv7hl or does armv5tel not need the ExclusiveArch tag?
Quoting DJ Delorie dj@redhat.com:
Based on an internal mirror of the various repositories (i.e. it might be slightly out of date), I wrote a script to compare the SRPMs (diff'ing the contents) generated during stage4 with SRPMs from other sources (mostly, F15 official ones, but also from SRPMs still in the builder queues). Color coded with diffs and stats:
http://djdelorie.fedorapeople.org/armv7-srpms.html
(uses javascript to show/hide diffs)
Summary:
99 SRPMs are of the .0.armN variety (only 5 I couldn't find the upstream to diff against)
3 SRPMs in stage 4 have the same NVR as GA/updates BUT DIFFER IN CONTENT (glibc and gdesklets vs GA, crash has changelog diffs vs development)
5 SRPMs match rawhide
11 SRPMs show up in the builder queues ("others") but not upstream
0 SRPMs that show up in the builder queues differ in content
30 SRPMs I couldn't find outside stage4 (might be from an older rawhide)
4704 SRPMs match GA or updates (name and content match)
Note: the HTML file is about 1 MB _______________________________________________ arm mailing list arm@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/arm
omalleys@msu.edu wrote:
Nice chart! I was poking through the diffs, and in a number of spots It shows:
-ExclusiveArch: %{ix86} x86_64 ppc alpha sparcv9 +ExclusiveArch: %{ix86} x86_64 ppc alpha sparcv9 armv7hl
I thought we were adding armv5tel as well when we added armv7hl or does armv5tel not need the ExclusiveArch tag?
Should we not simply add %{arm} and catch all variants?
d.marlin ========
Quoting DJ Delorie dj@redhat.com:
Based on an internal mirror of the various repositories (i.e. it might be slightly out of date), I wrote a script to compare the SRPMs (diff'ing the contents) generated during stage4 with SRPMs from other sources (mostly, F15 official ones, but also from SRPMs still in the builder queues). Color coded with diffs and stats:
http://djdelorie.fedorapeople.org/armv7-srpms.html
(uses javascript to show/hide diffs)
Summary:
99 SRPMs are of the .0.armN variety (only 5 I couldn't find the upstream to diff against)
3 SRPMs in stage 4 have the same NVR as GA/updates BUT DIFFER IN CONTENT (glibc and gdesklets vs GA, crash has changelog diffs vs development)
5 SRPMs match rawhide
11 SRPMs show up in the builder queues ("others") but not upstream
0 SRPMs that show up in the builder queues differ in content
30 SRPMs I couldn't find outside stage4 (might be from an older rawhide)
4704 SRPMs match GA or updates (name and content match)
Note: the HTML file is about 1 MB _______________________________________________ arm mailing list arm@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/arm
tor 2011-09-29 klockan 09:27 -0400 skrev omalleys@msu.edu:
Nice chart! I was poking through the diffs, and in a number of spots It shows:
-ExclusiveArch: %{ix86} x86_64 ppc alpha sparcv9 +ExclusiveArch: %{ix86} x86_64 ppc alpha sparcv9 armv7hl
I thought we were adding armv5tel as well when we added armv7hl or does armv5tel not need the ExclusiveArch tag?
There is about 100 of those related to Haskell. Hopefully we will manage to get ghc running on armv5tel as well but it can't be done until a fair bit of f15 base system is there. As ghc is itself written in haskell there is a bit of procedure to get it built per arch.
Regards Henrik