[FFmpeg-devel] [PATCH] Use pkgconfig for dirac libs

Måns Rullgård mans
Thu Oct 23 13:23:46 CEST 2008


Luca Barbato wrote:
> On 23-10-2008 12:52, M?ns Rullg?rd wrote:
>> As I said, different.  Additionally, pkgconfig searches according to
>> PKG_CONFIG_PATH (or similar), whilst the linker uses LIBRARY_PATH.
>> If these differ for whatever reason, you'll end up with a mismatch.
>
> ...

What is that supposed to mean?  That you don't care?

>>>> Furthermore, it is much less work to add a fairly trivial bit of code
>>>> to half a dozen linkers (and pkgconfig only works with gnu tools
>>>> anyway)
>>> and microsoft bad excuse for compiler (aka cl)...
>>
>> Do I look like I care?  I'm talking about the native compilers on
>> various Unix systems and cross-compilers from non-gnu vendors.  These
>> are often much better than gcc.
>
> fdo has a bugzilla as well (btw which of them isn't compatible with the
> gcc syntax for what is important here?)

GCC has numerous flags that are not compatible with other compilers.  It
is easier to list the ones that *are* compatible across a wide range.
The -D, -E, -I, -L, -O, -U, -c, -g, -l, and -o flags are all standard flags
specified by POSIX, and most compilers seem to support them, even when not
otherwise POSIX compatible.  The standard also includes -s (to produce a
stripped output), but not many compilers implement this (or even use -s
to mean something else).

Examples of common gcc flags *not* supported by other compilers are
all the -f, -m, and -W flags.

>>>> , than it is to modify each and every package ever written to
>>>> use pkgconfig (even if we pretend for a moment that using it wasn't
>>>> associated with severe mental suffering).
>>> s/modify/replace at least 10 lines of cruft per dep with a single one
>>> iterable over a list/
>>>
>>> Please spend a bit of time checking...
>>
>> Gentoo has well over 10000 packages.  If we want to add a feature to
>> 10 linkers, it would take 1000 lines of code in each for this task to
>> match that of updating all packages.  I also believe that investing
>> a little more time, even 10 times more would acceptable, upfront in
>> fixing the root problem (static libs not recording dependencies) would
>> mean less work in the longer term.  This gives us 10000 lines of code
>> to implement the new linker feature, which I'm certain is more than
>> enough.  We must also not forget to take into account the burden of
>> writing and maintaining pkgconfig itself.
>
> If is that useful and easy to archive why nobody even proposed that before?

Maybe because the original problem isn't as bad as it's made out to be.

-- 
M?ns Rullg?rd
mans at mansr.com




More information about the ffmpeg-devel mailing list