[Buildroot] [PATCH v2 1/1] libglib2: bump to 2.58.3

[Arnout Vandecappelle]

On 05/02/2019 20:29, Thomas Petazzoni wrote:
> On Tue,  5 Feb 2019 19:56:23 +0100
> aduskett at gmail.com wrote:
> 
>> From: Adam Duskett <Aduskett at gmail.com>
>>
>> In addition:
>>  - Update second patch
>>  - Remove third and fifth patches (already in version)
>>  - Add a new patch to fix a missing header
>>  - Add LIBGLIB2_GTK_DOC_HOOK so autoreconf do not fail on the following
>>    error:
>>      automake: error: cannot open < gtk-doc.make: No such file or directory
>>
>>  - Add a new patch: package/libglib2/0005-use-tooldir-in-pc-file.patch
>>    This patch fixes the previous autobuild errors by modifying the glib-2.0.pc
>>    file and adding a tooldir variable, of which the glib_genmarshal,
>>    gobject_query, and glib_mkenums of which will be prefixed. Then a
> 
> I am confused by the double "of which" usage in this sentence.
> 
>> +define LIBGLIB2_FIX_PC_FILE
>> +$(SED) "s at toolsdir=.*@toolsdir=$(STAGING_DIR)/usr/bin at g" \
>> +	$(STAGING_DIR)/usr/lib/pkgconfig/glib-2.0.pc
> 
> This kind of replacement with STAGING_DIR is a bit annoying for the
> upcoming per-package stuff. Options that I see:

 Why is it annoying? It will point to libglib2's staging dir, but that's fine,
isn't it?

> 
>  (1) Add 'bindir' to the list of pkg-config variables that should be
>      sysroot-prefixed. Works, but impact unknown.

 With this option, the patch is not even needed!

 Unfortunately, it might result in runtime breakage too... if some .pc's bindir
is used to set the path to some executable... Like

CFLAGS += -DBASH_PATH=`pkg-config bash --variable=bindir`/bash


>  (2) Add 'toolsdir' to the list of pkg-config variables that should be
>      sysroot-prefixed. Works, but impact unknown.

 Something similar was in fact my plan, that in the pkg-config wrapper we add
--define-variable=toolsdir=$STAGING_DIR

 Regards,
 Arnout

> 
>  (3) Make toolsdir=${libdir}/../bin. Works, but ugly.
> 
> I don't have a good opinion on what is the best choice here. Let's see
> what Arnout says.
> 
> Thomas
>