build/pkgs/ccache: Update to 4.10.2

[gmou3]

I also modified sdh_cmake slightly, because inputting . and .. together issued a warning.

Resolves #35872.

[github-actions]

Documentation preview for this PR (built with commit 02273ef; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[kwankyu]

I tested the branch (and the new ccache) on a mac. Did you test it on any linux?

[gmou3]

Yes, seems to work fine.

[kwankyu]

OK. LGTM.

Thanks.

[enriqueartal]

I had an issue in Fedora 41, ccache installation did not work having some issue with Xxhash. It worked after installing system package xxhash-devel; should this dependency appeared somewhere? On the other side system package ccache is installed and a fedora.txt exists in the corresponding distro folder, but it is ignored.

[kwankyu]

See https://groups.google.com/g/sage-devel/c/ZZIvoEHm5hg for the issue.

[dimpase]

Why does this PR touch a lot a files not related in any way to ccache?

[dimpase]

Another problematic change is that ccache now needs xxhash, and it tries to download it from the internet - something that probably breaks the rules for optional packages. And it also is flaky, see
https://groups.google.com/d/msgid/sage-release/CAD0p0K51VNkbovm1De3arXt18E9rpoP%2B%3DWO1414Cb-b_FP0%3D6Q%40mail.gmail.com

[gmou3]

Why does this PR touch a lot a files not related in any way to ccache?

I also modified sdh_cmake to not automatically run cmake in the current directory. This allows us to follow a package's installation instructions as recommended by the author (e.g., possibly cmake ..).

Another problematic change is that ccache now needs xxhash, and it tries to download it from the internet - something that probably breaks the rules for optional packages. And it also is flaky, see https://groups.google.com/d/msgid/sage-release/CAD0p0K51VNkbovm1De3arXt18E9rpoP%2B%3DWO1414Cb-b_FP0%3D6Q%40mail.gmail.com

Indeed, this is an issue. See #39226 for the proposed solution.

[dimpase]

The main issue with this PR is that now for some people installation of the optional package ccache is triggered without them asking for it! This is a bug.

Please see e.g. https://groups.google.com/g/sage-release/c/muFP_3aonHc/m/014QJ6j9DQAJ
and other messages in https://groups.google.com/g/sage-release/c/muFP_3aonHc/m/i7P0R-o_AQAJ

[gmou3]

That's unexpected, its optional status hasn't changed.
I'll try a clean build on my machine later to see if I can reproduce the issue (did you try this by any chance?).
I can't post on the discussion you shared, but it seems that one of the two users has a problem relevant to ccache while the other does not.

[dimpase]

That's unexpected, its optional status hasn't changed. I'll try a clean build on my machine later to see if I can reproduce the issue (did you try this by any chance?). I can't post on the discussion you shared, but it seems that one of the two users has a problem relevant to ccache while the other does not.

no, at least 2 users, @JohnCremona and @EmmanuelCharpentier report unwanted installation of ccache. It could be they run same or similar versions of Ubuntu.

[EmmanuelCharpentier]

Le jeudi 23 janvier 2025 à 10:58 -0800, Dima Pasechnik a écrit :

> That's unexpected, its optional status hasn't changed. I'll try a > clean build on my machine later to see if I can reproduce the issue > (did you try this by any chance?). I can't post on the discussion > you shared, but it seems that one of the two users has a problem > relevant to ccache while the other does not. no, at least 2 users, @JohnCremona [1] and @EmmanuelCharpentier [2] report unwanted installation of ccache. It could be they run same or similar versions of Ubuntu.

I'm running Debian testing, upgraded (more or less) daily.

-- Emmanuel Charpentier

— Reply to this email directly, view it on GitHub [3], or unsubscribe [4]. You are receiving this because you were mentioned.Message ID: ***@***.***>

[1] @JohnCremona https://github.com/JohnCremona [2] @EmmanuelCharpentier https://github.com/EmmanuelCharpentier [3] view it on GitHub #39119 (comment) [4] unsubscribe https://github.com/notifications/unsubscribe-auth/AB4WGTH2MHKKFAUMLMKSBB32ME3T5AVCNFSM6AAAAABTMYOBS6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMJQG43TSOJRG4

[JohnCremona]

My machine runs Ubuntu 22

[dimpase]

Why does this PR touch a lot a files not related in any way to ccache?

I also modified sdh_cmake to not automatically run cmake in the current directory. This allows us to follow a package's installation instructions as recommended by the author (e.g., possibly cmake ..).

I don't think it's needed any more. Early versions of cmake required a separate build directory, and these instructions might be a leftover from these early days.

Did you try without making these changes?

---

I don't see this problem (of unrequested building of ccache spkg) on Gentoo, so this makes for an "interesting" platform-specific bug...

[gmou3]

Why does this PR touch a lot a files not related in any way to ccache?

I also modified sdh_cmake to not automatically run cmake in the current directory. This allows us to follow a package's installation instructions as recommended by the author (e.g., possibly cmake ..).

I don't think it's needed any more. Early versions of cmake required a separate build directory, and these instructions might be a leftover from these early days.

Did you try without making these changes?

cmake requires a build directory, and often .. is used. Not only ..

Not making such changes would require me to diverge from the recommended ccache instructions, but I wanted to avoid the uncertainties of this. I think this makes the translation of build instructions easier and clearer for all packages. cmake becomes sdh_cmake and that's it.

I don't see this problem (of unrequested building of ccache spkg) on Gentoo, so this makes for an "interesting" platform-specific bug...

So it probably isn't a result of this PR. Maybe the failure just made the building of ccache apparent.

[dimpase]

So it probably isn't a result of this PR. Maybe the failure just made the building of ccache apparent.

@JohnCremona - could you please check whether reverting this PR (i.e. the commits at https://github.com/sagemath/sage/pull/39119/commits) fixed the issue of the unasked for build of ccache ?

[JohnCremona]

So it probably isn't a result of this PR. Maybe the failure just made the building of ccache apparent.

@JohnCremona - could you please check whether reverting this PR (i.e. the commits at https://github.com/sagemath/sage/pull/39119/commits) fixed the issue of the unasked for build of ccache ?

Sorry for the delay, I was away. I will try.

By reverting the PR, do you mean checking out the commit before the first of these and building from scratch?

I did solve the problem by unsetting that macro, so is it necessary to do more?