mintro: dump ast of subdir meson.build files

[bruchar1]

No description provided.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (dea72e4) 70.30% compared to head (019a0e2) 68.83%.

❗ Current head 019a0e2 differs from pull request most recent head 384e8b0. Consider uploading reports for the commit 384e8b0 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #11446      +/-   ##
==========================================
- Coverage   70.30%   68.83%   -1.48%     
==========================================
  Files         219      209      -10     
  Lines       48478    45560    -2918     
  Branches    11461     9433    -2028     
==========================================
- Hits        34082    31360    -2722     
+ Misses      11963    11777     -186     
+ Partials     2433     2423      -10     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[bruchar1]

Oups! I forgot to include ast from the root meson.build... Should be better now.

[xclaesse]

Oups! I forgot to include ast from the root meson.build... Should be better now.

Interesting, I would have expected the root meson.build to be in self.processed_buildfiles, but it's not. The real interpreter does that too, which has the effect that subdir('.') in the root meson.build file parse itself again but luckily abort with ERROR: Second call to project().

[xclaesse]

Now the question is does backward compat matter? I have no idea who uses --ast... Maybe would need to be guarded with --recursive flags? dunno.

[bruchar1]

Now the question is does backward compat matter? I have no idea who uses --ast... Maybe would need to be guarded with --recursive flags? dunno.

I doubt it could be used for anything else that debugging... Not dumping subdirs means the result was unusable except for projects with only one meson.build file. Imho, that should not break anything important. If something breaks with this, it would break with any change in the structure of the AST...

[bruchar1]

Interesting, I would have expected the root meson.build to be in self.processed_buildfiles, but it's not.

Hum... I changed it. This is the right thing to do. It simplifies the dump_ast function.

[bruchar1]

@xclaesse Was everything ok with you?

[xclaesse]

If something breaks with this, it would break with any change in the structure of the AST...

Not really, because this changes the output to a dict, so any tool that was parsing this json are now broken, if any. Wondering if the CLI syntax that requires passing meson.build argument should mean that we want AST for only that particular file. And if we change to allow passing a directory then that would mean recursive.

[bruchar1]

Not really, because this changes the output to a dict, so any tool that was parsing this json are now broken, if any.

Technically, the output is still json... It just is one level deeper.

Wondering if the CLI syntax that requires passing meson.build argument should mean that we want AST for only that particular file. And if we change to allow passing a directory then that would mean recursive.

That could be great. However, this would require #11485, changes to the introspection interpreter to be able to parse a meson file without project node, and changes to the subdir visitor.

[bruchar1]

I just updated this PR. Instead of breaking the output of the existing --ast, I introduced a new --full-ast option to introspect.