Add @BASENAMES@ and @PLAINNAMES@ to custom_target

[dcbaker]

There are currently projects in the wild (I don't want to shame them), who are recommending this:

custom_target(
  ...
  input : ['a.in', 'b.in', 'c.in'],
  output : '.'
  command : [find_program('cmd'), '@OUTPUT@', '@INPUT@'],
)

of course, this isn't what they actually want, they just don't want to have to list ['a.out', 'b.out', 'c.out'] in their output lines. So lets make it easy to do the right thing:

custom_target(
  ...
  input : ['a.in', 'b.in', 'c.in'],
  output : '@[email protected]',
  command : [find_program('cmd'), 'batch', '@OUTDIR@', '@INPUT@'],
)

With the plural form passed (it is allowed only to be passed as an exclusve output argument), meson will replace that with the @PLAINNAME@ applied to each output, making it simple to handle a custom_target with multiple outputs that follow the @BASENAME@ or @PLAINNAME@ scheme.

still needs tests, but I figure we're going to bikeshed about the naming for a while, so...

[codecov[bot]]

Codecov Report

Attention: Patch coverage is 39.13043% with 14 lines in your changes missing coverage. Please review.

Project coverage is 43.92%. Comparing base (ee7a7fe) to head (258a2fb). Report is 2195 commits behind head on master.

Files	Patch %	Lines
mesonbuild/mesonlib/universal.py	42.85%	6 Missing and 2 partials :warning:
mesonbuild/interpreter/interpreter.py	33.33%	4 Missing and 2 partials :warning:

:exclamation: There is a different number of reports uploaded between BASE (ee7a7fe) and HEAD (258a2fb). Click for more details.

HEAD has 16 uploads less than BASE

Flag	BASE (ee7a7fe)	HEAD (258a2fb)
	19	3

Additional details and impacted files

@@             Coverage Diff             @@
##           master   #10498       +/-   ##
===========================================
- Coverage   68.74%   43.92%   -24.83%     
===========================================
  Files         406      406               
  Lines       87868    87485      -383     
  Branches    19530    18554      -976     
===========================================
- Hits        60409    38425    -21984     
- Misses      22886    45402    +22516     
+ Partials     4573     3658      -915     

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[eli-schwartz]

of course, this isn't what they actually want, they just don't want to have to list ['a.out', 'b.out', 'c.out'] in their output lines. So lets make it easy to do the right thing:

Actually, these projects just want to place output files in subdirectories. It's not technically required, but the transpiler cmd in question simultaneously documents the meson snippet using output: '.' and makes the tool emit the files in directory structure. The example project using the cmd, then makes use of file aliases in the compiler to expose all those files without a directory structure at all.

In theory, the real solution here is generator(...).process(..., preserve_path_from: meson.current_source_dir()), in practice generators are totally useless unless you're outputting *.c *.cpp *.h.

[dcbaker]

I thought more about this, I think instead of adding a new @BASENAMES@ template, allowing @BASENAME@ to contain the array of outputs like @INPUT@ and @OUTPUT@ do would be more consistant and natural, which would also allow for @BASENAME1@, which is probably useful. This shouldn't cause any regressions because currently passing @BASENAME@ or @PLAINNAME@ with multiple outputs is an error.