Delete Percona-Server/CMakeCache.txt from the build-binary.sh script

[Roel Van de Paar]

When building various builds (./build/build-binary.sh . ; ./build/build-binary.sh --debug . ; ./build/build-binary.sh --debug --valgrind .) it's easy to forget to include deletions of Percona-Server/CMakeCache.txt

This feature request to see if we can/should add 'rm Percona-Server/CMakeCache.txt' into the build-binary.sh script so that it's a cleaner build automatically, each time.

(hijack; at the same time, can we also make it so that if no options are passed, "." is assumed as default $1?)

[Stewart Smith (stewart)]

RoelV <email address hidden> writes:
> Public bug reported:
>
> When building various builds (./build/build-binary.sh . ; ./build/build-
> binary.sh --debug . ; ./build/build-binary.sh --debug --valgrind .)
> it's easy to forget to include deletions of Percona-
> Server/CMakeCache.txt
>
> This feature request to see if we can/should add 'rm Percona-
> Server/CMakeCache.txt' into the build-binary.sh script so that it's a
> cleaner build automatically, each time.

This is probably a really good idea, CMake is fairly awful at doing such
things, and as a workaround we have the bzr clean-tree option enabled in
most Jenkins builds.

--
Stewart Smith

[Laurynas Biveinis (laurynas-biveinis)]

How about making these scripts work on empty directories only?

[Roel Van de Paar (roel11)]

If the CMakeCache.txt deletion is considered enough to ensure clean builds, that would be an easy fix & good enough for QA. Is it enough though?

[Laurynas Biveinis (laurynas-biveinis)]

Enforcing out-of-source builds on empty directories only would nip any stale build artifact issue in the bud, but maybe removing CMakeCache.txt is good enough, so I am not sure.

[Ignacio Nin (ignacio-nin)]

Hi,

The binary scripts, unlike their rpm/dpkg counterparts, configure and make in the source directory. This is inconvenient for many reasons, including the one that inspired this bug.

I was working on a solution to have it configure to another directory (perhaps export the source like the other scripts) so the source directory remains unchanged. This should cover this issue and some others.

I'll link that branch to this bug, let me know if this is an acceptable solution.

About the target directory, already the default directory is .

[Stewart Smith (stewart)]

Laurynas Biveinis <email address hidden> writes:
> Enforcing out-of-source builds on empty directories only would nip any
> stale build artifact issue in the bud, but maybe removing CMakeCache.txt
> is good enough, so I am not sure.

out of source builds is a good idea anyway, it appears to be the most
supported way going forward for MySQL.

I'd prefer it if we even did out of source builds after generating
source tarball.

--
Stewart Smith

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PS-1352