Re: [Gimp-print-devel] PKG_LIB_DIR/PKGMODULEDIR expansion issues

[Matsumura, G. <gm9...@oh...>]

Greetings,

In order to solve the mentioned issues, I was able to create the
attached prefix_defines.patch which moves the expansion of the ${prefix}
and ${exec_prefix} variables to build time through make as opposed to
configure time through autoconf. This is in accordance with the
suggestion here:
https://www.gnu.org/software/autoconf/manual/autoconf-2.67/html_node/Installation-Directory-Variables.html

Specifically, this guideline was followed:

    "In order to support these features, it is essential that
    datarootdir remains defined as ‘${prefix}/share’, so that its value
    can be expanded based on the current value of prefix.

    A corollary is that you should not use these variables except in
    makefiles. For instance, instead of trying to evaluate datadir in
    configure and hard-coding it in makefiles using e.g.,
    ‘AC_DEFINE_UNQUOTED([DATADIR], ["$datadir"], [Data directory.])’,
    you should add -DDATADIR='$(datadir)' to your makefile's definition
    of CPPFLAGS (AM_CPPFLAGS if you are also using Automake)."

This approach is somewhat similar to that which existed in gutenprint
before commit b6043e. I wasn't sure if there was a particular reason why
this method was undesirable/removed at the time.

As a consequence of my build of gutenprint not being able to find the
modules, I hit the segfault described here:
https://sourceforge.net/p/gimp-print/bugs/732/
The attached no_modules_segfault.patch alters this behaviour so that
stp_printer_model_count is called instead of stp_list_get_length,
leading to an informative warning being printed in the case of no
modules instead of a segmentation fault. This is consistent with usage
in the rest of the code.

Thank you for reading this, and any consideration of these changes would
be very much appreciated. I apologize for any mistakes I made in
creating the patches. If there is anything I can do to improve them,
please let me know.

Regards,
    George

On 7/1/22 21:30, Matsumura, George wrote:
> Greetings,
> 
> My build of Gutenprint, built through Void Linux's standardized build
> process, was unable to find the module directories due to PKGMODULEDIR
> being defined in config.h as:
> #define PKGMODULEDIR "${exec_prefix}/lib64/gutenprint/5.3/modules"
> 
> Since the C processor cannot expand exec_prefix, it remains in the path
> at runtime and cannot be resolved.
> 
> Attached is the config.summary from the build. In particular, the
> following flag passed to configure seemed to be responsible:
> --libdir=\${exec_prefix}/lib64
> 
> There seems to be a specific case in configure.ac for when libdir is
> defined as \${exec_prefix}/lib, but this would not cover my case.
> 
> This seems to be similar to the problem described here:
> https://www.gnu.org/savannah-checkouts/gnu/autoconf/manual/autoconf-2.71/html_node/Defining-Directories.html
> 
> Is there a recommended way to patch configure.ac so that such an option
> works correctly? The above reference suggests either moving definitions
> to compile-time options in Makefile.am, which, looking at the repository
> history, seems to have been done in gutenprint in the past, or using an
> additional macro to resolve the pathname.
> 
> Thank you for reading this and in advance for any reply.
> 
> Regards,
>     George