[question] Using full_package_mode for glib with gtk dependencies causes brittle builds

[theartful]

My PR for building gtk on windows was successfully built #11308, but after #11545 it can't be built anymore, since harfbuzz updated its glib version. And since gtk and its dependencies use full_package_mode for glib (which is necessary to prevent multiple instances of glib), I get MISSING_DEPENDENCIES error for harfbuzz (as harfbuzz with the old glib is not in the cache anymore).

This hard dependency makes it hard to update glib, since now if glib's version is bumped in any of gtk's dependencies, gtk will fail to build in the CI until all other dependencies bump glib's version as well.

How do you think this should be handled?

[planetmarshall]

I'm not sure what the best solution for this is - I'm inclined to suggest static glib should be an unsupported configuration (I can't debug a GTK app in Linux built with static glib - even with everything built from scratch), but I'm not sure that would solve the problem.

This is really an issue with the cached binaries, as there's nothing really wrong with the recipe - it can be built using --build missing by the end user with no problems.

[morevnaproject]

Hi! Any news on this topic? It would be really nice to have a solution, as we need Gtk. ^__^

[planetmarshall]

I think this is going to continue to be a problem for any recipes where full_package_mode is specified. Is it possible for CCI packages to be built with --build missing? Or would that just consume too many resources?

[uilianries]

CCI will not use --build missing. Will need to continue to update recipes.

[theartful]

Now that glib is shared, I think we should be removingfull_package_mode (or at least use it only when linking against static glib). These patches are really causing problems.

[uilianries]

@hesham-essam Yes, removing full_package_mode sounds correct. We will need new PRs with those fixes.

[theartful]

This is solved now