[CMakeToolchain][env] Added PKG_CONFIG_PATH env variable and PKG_CONFIG_EXECUTABLE one

[franramirez688]

Changelog: Feature: Added generators folder to PKG_CONFIG_PATH environment variable in CMakeToolchain.
Changelog: Feature: Ensure that CMakeToolchain will enforce usage of pkg-config executable set in tools.gnu:pkg_config config.
Closes: #11962
Partial implementation of #12354
Docs: conan-io/docs#2832

Note: Originally started by #12355

[franramirez688]

Should I use this approach from EnvVars?

        self._subsystem = deduce_subsystem(conanfile, scope)

    @property
    def _pathsep(self):
        return ":" if self._subsystem != WINDOWS else ";"

[czoido]

Yes, I think the correct way to do this is to deduce the subsystem and then aplying the path separator according to that

[franramirez688]

The question is: should I assume that scope == "build"? Here, we don't have any scope defined.

[memsharded]

Nice job with this test!

[jcar87]

I like this! :D Matches the documentation - and still allows (a few lines above in the template block) to fall back on anything the user may have previously specified

PKG_CONFIG_PATH

A colon-separated (on Windows, semicolon-separated) list of directories to search for .pc files. The default directory will always be searched after searching the path; the default is libdir/pkgconfig:datadir/pkgconfig where libdir is the libdir where pkg-config and datadir is the datadir where pkg-config was installed.

[hoyhoy]

I'm running conan version 1.60.1 and I'm seeing self.conf.get("tools.gnu:pkg_config", check_type=str) == None when my conan.conf is set with the following.

[tools]
gnu:pkg_config = /usr/local/bin/pkgconf

Also, if I set pkg_config with the conan config command, it corrupts conan.conf.

conan config set tools.gnu:pkg_config=/usr/local/bin/pkgconf

gnu = pkg_config = /usr/local/bin/pkgconf

It only seems to work if I put a [conf] section in my conan profile.

[memsharded]

Hi @hoyhoy

This is a closed issue from 8 months ago, it would be better to report this as a new ticket.
There are some known issues, if the pkgconf is called at package_info() method it will not work, because environment or conf is not injected that way at that point.

Also [tools] is not a valid section, and conan.conf is not the right place to put it (in any case, it would be global.conf. Likewise gnu:pkg_config is not a conf, but it must be tools.gnu:pkg_config=value, the full string, and in global.conf

[hoyhoy]

Hmmm, conan config list doesn't refer to conan.conf? conan config set absolutely changes that file. Also, self.conf.get("tools.gnu:pkg_config", check_type=str) doesn't refer to conan.conf? If it's global.conf, why does conan config write to conan.conf? I'm not sure how to write an issue. It definitely seems like something is wrong.

[memsharded]

Hmmm, conan config list doesn't refer to conan.conf? conan config set absolutely changes that file. Also, self.conf.get("tools.gnu:pkg_config", check_type=str) doesn't refer to conan.conf? If it's global.conf, why does conan config write to conan.conf?

No, you are mixing things. In Conan 1.X the two different confs coexist:

• conan.conf is the old, legacy one. conan config set/update writes to it, but those commands are also removed in Conan 2.0. It contains the legacy configuration items
• global.conf or profile files are the new ones. conan config list list the possible values, but they should go to global.conf or profiles. tools.gnu:pkg_config is a new configuration listed here, it should be defined in global.conf or profile.