enum exclusions not working with ffigen

[jardon]

hello!

im new to ffigen, but i think that ive hit a bug. i have been attempting to create bindings for a C library using ffigen and need to exclude enums as those need to be implemented in dart. according to the docs, the enums.exclude key should be used to pass exclusions to ffigen. this happens on 19.1.0 and on main. based on the docs and the ffigen schema the following snippet should work afaik:

name: FlatpakBindings
output: 'lib/src/flatpak_bindings.dart'
headers:
  entry-points:
    - '/usr/include/flatpak/flatpak.h'
compiler-opts:
  - '-I/usr/include/glib-2.0'
  - '-I/usr/lib/x86_64-linux-gnu/glib-2.0/include' # Or similar path for glibconfig.h
  - '-I/usr/include/gio-unix-2.0'
  - '-I/usr/include/flatpak'
enums:
  exclude:
    - 'FlatpakInstallFlags'

despite the config, i still get the following error:

[SEVERE] : The integer type used for enums is implementation-defined. FFIgen tries to mimic the integer sizes chosen by the most common compilers for the various OS and architecture combinations. To prevent any crashes, remove the enums from your API surface. To rely on the (unsafe!) mimicking, you can silence this warning by adding silence-enum-warning: true to the FFIgen config. Affected enums:
	FlatpakInstallFlags

[dcharkes]

If I remember correctly, the warning is emitted before the exclusion logic is run. We should have an open bug for that, but I can't find it.

[jardon]

im not sure if that is the same issue. in my testing, providing an exclude in the config versus not having one at all gives the same result and generates bindings with the enums

[jardon]

additionally, i was able to use as-int with an excludes and it processed the exclusions correctly for that context. the top level enums.excludes is what this bug report is about.

using as-int.excludes does, however, still give the enums warning SEVERE error from above so i think that warning order issue @dcharkes brought up is a separate, valid issue

[liamappelbe]

Is there a bug here aside from the spurious error message? Is the enum actually being code genned?

If so, I'd look for places in your API that use the enum. FFIgen will code gen transitive dependencies, so for example if you have a function that returns this enum, you'll probably get code generation for it, even if you try to exclude it. Then you'll have to decide how to deal with those APIs that transitively use this enum. Do you want to exclude them too?

[jardon]

i can confirm that this works as expected given a sample header file and config. yes, the message is still misleading. this issue can either be converted to track the message issue or closed as invalid -- maintainer's choice 😃

thanks!