Meson install warnings when running postgres build from a sandbox

From: "Matt Smith (matts3)" <matts3(at)cisco(dot)com>
To: "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org>
Subject: Meson install warnings when running postgres build from a sandbox
Date: 2025-10-22 00:57:47
Message-ID: MW6PR11MB8310D6607E09222C79D7B1B68CF3A@MW6PR11MB8310.namprd11.prod.outlook.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

If I run postgres builds from a bazel sandbox<https://bazel.build/docs/sandboxing#sandboxing-strategies> whereby the source directory consists of symlinks that point to the original source files.

Meson has traditionally followed these symlinks, but since 1.3<https://github.com/mesonbuild/meson/commit/0af126fec798d6dbb0d1ad52168cc1f3f1758acd>, any of the install_* functions produce a warning:

The [[install_data]], [[install_headers]], [[install_subdir]] functions now
have an optional argument `follow_symlinks` that, if set to `true`, makes it so
symbolic links in the source are followed, rather than copied into the
destination tree, to match the old behavior. The default, which is currently
to follow links, is subject to change in the future.

This is all ok for now, but I'm wondering if meson changes the default behavior in a future version, how postgres would be able to support the old way of following symlinks with said future version.

This would require a new build option in combination with setting follow_symlinks<https://mesonbuild.com/Release-notes-for-1-3-0.html#added-follow_symlinks-arg-to-install_data-install_header-and-install_subdir> to this option value, but I also noticed that postgres supports a minimum version of >=0.57.2 which does not have this arg on any of the install_* functions.

Thanks

Matt

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Bryan Green 2025-10-22 01:09:56 Re: [PATCH] Add Windows support for backtrace_functions (MSVC only)
Previous Message Joel Jacobson 2025-10-22 00:02:08 Re: LISTEN/NOTIFY bug: VACUUM sets frozenxid past a xid in async queue