We're trying to track down a problem that only appears in koji and copr, but I can't reproduce locally in mock on my rawhide VM. It's related to file triggers and we're trying to debug with something this:
%transfiletriggerin -n texlive-kpathsea -- /usr/share/texlive # %{_bindir}/texhash 2> /dev/null || : # DEBUG, lets see what it does /usr/bin/sh -x %{_bindir}/texhash || : export TEXMF=/usr/share/texlive/texmf-dist export TEXMFCNF=/usr/share/texlive/texmf-dist/web2c export TEXMFCACHE=/var/lib/texmf # %{_bindir}/mtxrun --generate &> /dev/null || : # %{_bindir}/fmtutil-sys --all &> /dev/null || : # DEBUG, lets see what it does %{_bindir}/mtxrun --generate || : %{_bindir}/fmtutil-sys --all || :
hoping to see the output in the mock root.log, but I get nothing.
One basic question - if texlive-kpathsea is part of the transaction, does the above trigger still fire at the end of the transaction?
I see that mock locally and in copr has output like:
Running scriptlet: texlive-base-7:20200327-4.fc33~bootstrap.1.x86_6 357/357
But I don't see any "Running triggers" or similar. Are we correct to expect output from triggers?
This is causing build failures after the recent texlive update. Any help would be greatly appreciated.
On 5/19/20 6:01 AM, Orion Poplawski wrote:
We're trying to track down a problem that only appears in koji and copr, but I can't reproduce locally in mock on my rawhide VM. It's related to file triggers and we're trying to debug with something this:
%transfiletriggerin -n texlive-kpathsea -- /usr/share/texlive # %{_bindir}/texhash 2> /dev/null || : # DEBUG, lets see what it does /usr/bin/sh -x %{_bindir}/texhash || : export TEXMF=/usr/share/texlive/texmf-dist export TEXMFCNF=/usr/share/texlive/texmf-dist/web2c export TEXMFCACHE=/var/lib/texmf # %{_bindir}/mtxrun --generate &> /dev/null || : # %{_bindir}/fmtutil-sys --all &> /dev/null || : # DEBUG, lets see what it does %{_bindir}/mtxrun --generate || : %{_bindir}/fmtutil-sys --all || :
hoping to see the output in the mock root.log, but I get nothing.
One basic question - if texlive-kpathsea is part of the transaction, does the above trigger still fire at the end of the transaction?
IIRC yes.
I see that mock locally and in copr has output like:
Running scriptlet: texlive-base-7:20200327-4.fc33~bootstrap.1.x86_6 357/357
But I don't see any "Running triggers" or similar. Are we correct to expect output from triggers?
Dnf output regarding scriptlets is not really useful beyond "its doing something" indication, at least here it seems to be merging multiple triggers into one line of output accounted for the installing/removing package rather than the trigger owner etc. If you can reproduce this directly with dnf, using '--rpmverbosity debug' should shed more light on what's going on, including actual scriptlet names and their owner packages.
This is causing build failures after the recent texlive update. Any help would be greatly appreciated.
It's entirely possible there its just another file trigger bug, it wouldn't be the first one uncovered by texlive. Please file a bug on rpm with reproducer info and lets work from there.
- Panu -
On 5/20/20 5:54 AM, Panu Matilainen wrote:
It's entirely possible there its just another file trigger bug, it wouldn't be the first one uncovered by texlive. Please file a bug on rpm with reproducer info and lets work from there.
- Panu -
Thanks for replying Panu. As near as we can tell now it was indeed just the texlive scriptlet not working properly. It seems to be okay now.