Re-run configure+make after installing an SPKG with sage -i
Closed this issue · 31 comments
This is an idea for fixing #27373.
As explained in comment#73 on that ticket, after installing a new package it is really necessary to re-run configure after installing a new package, whether it was an optional package that was not previously installed, or a standard package that was not previously installed (due to a configure-time check determining that it was not needed).
This is to ensure that files generated by configure reflect the new system configuration, in terms of what SPKGs are installed.
Further, it is advisable to re-run make all-build in case the sage -i installed packages whose dependents need to be rebuilt.
This also needs to ensure that if an SPKG is installed for some standard package, that we account for that when re-running configure: It must not try again to check the system for that dependency, and just reflect that the SPKG is installed. For example, from #27373, if Sage is first built using GMP/MPIR from the system, then we run sage -i mpir, the system configuration must be updated to reflect that we are now using the mpir SPKG instead of GMP/MPIR from the system. It's also necessary to rebuilt any of its dependent packages.
Depends on #27641
CC: @dimpase
Component: build
Author: Erik Bray
Branch/Commit: public/build/ticket-27642 @ e168732
Reviewer: Dima Pasechnik
Issue created by migration from https://trac.sagemath.org/ticket/27642
This branch mostly fixes the bug in #27373 for me.
As noted in the commit message this does not yet preserve arguments previously passed to configure. I think we can fix that by running ./config.status --config but I need to test that a bit more.
Last 10 new commits:
687cdf0 | Replace various --with-gmp=$SAGE_LOCAL flags in spkg-installs with a |
6de961d | Reworked this a bit more |
f76d6d5 | fix typo |
5f90ff7 | added a bit more explanation |
3dee97c | correct logic for SAGE_GMP_PREFIX etc |
0308142 | add the AX_ABSOLUTE_HEADER macro |
eea0685 | iml in particular is very picky about being given an absolute path to the |
6f29e86 | Add support for unconditional pre- and post-check actions in spkg-configure.m4 |
a2bab81 | Refactor the mpir/spkg-configure.m4 (and to a lesser extent the one for gmp) to use the new pre and post arguments to SAGE_SPKG_CONFIGURE |
a29056a | Trac #27642: Initial attempt at fix: re-run `configure` and `make all-build` after |
Author: Erik Bray
Branch: u/embray/build/ticket-27642
could you rebase it?
Changed branch from u/embray/build/ticket-27642 to public/build/ticket-27642
New commits:
191269e | Add support for unconditional pre- and post-check actions in spkg-configure.m4 |
f9c9a91 | Refactor the mpir/spkg-configure.m4 (and to a lesser extent the one for gmp) to use the new pre and post arguments to SAGE_SPKG_CONFIGURE |
248c071 | Trac #27642: Initial attempt at fix: re-run `configure` and `make all-build` after |
Trying this with ./configure leaves gmp not marked as an not to be installed, while mpir is marked as not to be installed. (to be fixed on #27641).
Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:
191269e | Add support for unconditional pre- and post-check actions in spkg-configure.m4 |
1c24df1 | Refactor the mpir/spkg-configure.m4 (and to a lesser extent the one for gmp) to use the new pre and post arguments to SAGE_SPKG_CONFIGURE |
dc8fd7f | Refactor the mpir/spkg-configure.m4 (and to a lesser extent the one for gmp) to use the new pre and post arguments to SAGE_SPKG_CONFIGURE |
4bce1a3 | Trac #27642: Initial attempt at fix: re-run `configure` and `make all-build` after |
Reviewer: Dima Pasechnik
now this branch leaves gmp not marked as an not to be installed, while mpir is marked as not to be installed, despite being fixed on #27641...
Your rebasing did not apparently include the fixes I made to #27641 (compare gmp/spkg-configure.m4)
gmp/spkg-configure.m4 was not touched on #27641, have a look at git show 1c24df1.
`
More precisely, you overwrote this change in #27641 comment:15
Right. That was intentional. It should not be changed.
Replying to @embray:
Right. That was intentional. It should not be changed.
Should I re-create the un-rebased branch and try again, or it's simply removing the changes to gmp/spkg-configure.m4 that would do?
Branch pushed to git repo; I updated commit sha1. New commits:
18d612d | leave GMP's spkg-configure.m4 as it was |
This is a bit of a mess. I just need to rebase on top of the final version of #27641, not add spurious new commits.
Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:
d9bbdd4 | Add support for unconditional pre- and post-check actions in spkg-configure.m4 |
a360518 | Refactor the mpir/spkg-configure.m4 (and to a lesser extent the one for gmp) to use the new pre and post arguments to SAGE_SPKG_CONFIGURE |
e168732 | Trac #27642: Initial attempt at fix: re-run `configure` and `make all-build` after |
Updated to the final version of #27641. Still need to kick around a bit with making sure it can preserve previous configure flags.
Too late, a previous version of this branch was released in 8.8.beta5
I am afraid it has to be reverted. It's not good as it is now.
I'm confused. How was it merged? It was not set to positive review, and you never replaced the branch with the merge commit as in your usual process?
Replying to @embray:
I'm confused. How was it merged? It was not set to positive review,
it was, see comment:14
I gather it wasn't merged in the usual way as it didn't come with a (formally) needed configure bump
I can confirm that this branch is merged in 8.8.rc0.
So this is done and dusted.