Bug#663902: ghostscript: FTBFS on hurd-i386: extra gs_realloc in symbols file

[Jonathan Nieder <jrnieder@gmail.com>]

Hi,

Pino Toscano wrote:

> [Date: Thu, 10 Sep 2009 19:59:44 +0200]

Please reset your clock. :)

[...]
> as shown on [1], ghostscript fails to build on GNU/Hurd, because of a symbol
> specified in the symbols file which is not compiled on GNU/Hurd.
> The reason is basically the implementation of the gs_realloc() function: in
> base/malloc_.h there is:
> [...]
> /* (At least some versions of) Linux don't have a working realloc.... */
> #ifdef linux
> #  define malloc__need_realloc
> void *gs_realloc(void *, size_t, size_t);
> #else
> #  define gs_realloc(ptr, old_size, new_size) realloc(ptr, new_size)
> #endif

It seems unlikely that ghostscript's assumptions about glibc on Linux
would fail while they would succeed on Hurd, so I would prefer a fix
that just gets rid of the "#ifdef linux" block.

Based on the vcs history, the gs_realloc hack is from 1998 or earlier,
so it might be from "Linux libc" days.

What do you think?

Thanks,
Jonathan