do @ variable substitution when parsing cmake format

[Jan200101]

the upstream behavior of configure_file in cmake parses both @var@ and ${VAR} and only supports disabling the bracket variables through the @ONLY argument, meson needs to follow this behavior for compatibility

this is an ugly solution but I think its best considering the alternatives:

• calling do_conf_str twice with both formats
  • will break if the first iteration places substitutes that match the format of the second iteration
• adapt the existing bracket format regex to support @
  • the existing regex is unholy and broken in more ways than I am willing to list, it would be better to completely gut this part of the code and replace it with regular python code
• rewrite the code to not use regex
  • this would be quite complex and I am not familiar enough with mesondefine to be able to find any regressions. The lack of specification on how its suppose to work does not help either

[dcbaker]

Thanks for looking at this, I had noticed this but never had time to fix it.

It would be really good to get test coverage for this in the test cases/common/14 configure file test

[Jan200101]

rebasing to un-screw CI
EDIT: it ended up becoming worse, somehow, I don't think this is caused by this PR

[eli-schwartz]

rebasing to un-screw CI
EDIT: it ended up becoming worse, somehow, I don't think this is caused by this PR

mkenums/main6.c:8:42: error: ‘MESON_TYPE_THE_XENUM’ undeclared (first use in this function)

This is not a known or preexisting failure. I would venture to say something about this PR is in conflict with behavior that that test case was relying on, but I haven't had a chance to review the implementation.

[eli-schwartz]

We have a file, enums.h.in:

meson/test cases/frameworks/7 gnome/mkenums/enums.h.in

Lines 7 to 21 in 8156e12

	G_BEGIN_DECLS
	/*** END file-header ***/
	/*** BEGIN file-production ***/
	/* enumerations from "@basename@" */
	/*** END file-production ***/
	/*** BEGIN value-header ***/
	GType @enum_name@_get_type(void) G_GNUC_CONST;
	#define @ENUMPREFIX@_TYPE_@ENUMSHORT@ (@enum_name@_get_type())
	/*** END value-header ***/
	/*** BEGIN file-tail ***/
	G_END_DECLS

We convert it to enums6.h.in and then use it as a template for glib-mkenums:

meson/test cases/frameworks/7 gnome/mkenums/meson.build

Lines 143 to 155 in 8156e12

	# Simple trick to copy the file without substitutions, can be
	# removed when https://github.com/mesonbuild/meson/pull/3383 is fixed
	gen_h_template = configure_file(input: 'enums.h.in',
	output: 'enums6.h.in',
	configuration: configuration_data(),
	format: 'cmake')
	enums_h6 = gnome.mkenums('enums6',
	sources : 'meson-sample.h',
	h_template : gen_h_template,
	ftail : '/* trailing header file info */',
	install_header : true,
	install_dir : get_option('includedir'))

The replacement tokens must be parsed by glib-mkenums, not configure_file. We use configure_file in order to copy a file without modifications, and that's what this PR causes to stop behaving the way it used to behave. If you configure the test case before and after this PR:

File created by glib-mkenums:

diff --git a/builddir-old/mkenums/enums6.h b/builddir-new/mkenums/enums6.h
index 313d846ff..950175623 100644
--- a/builddir-old/mkenums/enums6.h
+++ b/builddir-new/mkenums/enums6.h
@@ -8,11 +8,11 @@
 
 G_BEGIN_DECLS
 
-/* enumerations from "meson-sample.h" */
-GType meson_the_xenum_get_type(void) G_GNUC_CONST;
-#define MESON_TYPE_THE_XENUM (meson_the_xenum_get_type())
-GType meson_the_flags_enum_get_type(void) G_GNUC_CONST;
-#define MESON_TYPE_THE_FLAGS_ENUM (meson_the_flags_enum_get_type())
+/* enumerations from "" */
+GType _get_type(void) G_GNUC_CONST;
+#define _TYPE_ (_get_type())
+GType _get_type(void) G_GNUC_CONST;
+#define _TYPE_ (_get_type())
 
 G_END_DECLS
 

File created by configure_file:

diff --git a/builddir-old/mkenums/enums6.h.in b/builddir-new/mkenums/enums6.h.in
index 479867fd1..7f6a50107 100644
--- a/builddir-old/mkenums/enums6.h.in
+++ b/builddir-new/mkenums/enums6.h.in
@@ -9,11 +9,11 @@ G_BEGIN_DECLS
 
 /*** BEGIN file-production ***/
 
-/* enumerations from "@basename@" */
+/* enumerations from "" */
 /*** END file-production ***/
 /*** BEGIN value-header ***/
-GType @enum_name@_get_type(void) G_GNUC_CONST;
-#define @ENUMPREFIX@_TYPE_@ENUMSHORT@ (@enum_name@_get_type())
+GType _get_type(void) G_GNUC_CONST;
+#define _TYPE_ (_get_type())
 /*** END value-header ***/
 
 /*** BEGIN file-tail ***/

[eli-schwartz]

I think what this means is that the existing testcase code was relying on format: 'cmake' not actually implementing cmake replacements, and obviously implementing that missing feature meant that we then mangled the file by, well, copying it.

Or something.

Those @VARNAME@ tokens weren't part of the empty configuration_data used, obviously, so they ended up being deleted.