[MPlayer-users] Release 1.1 first packaging attempt

[Raimund Steger]

On Tue, May 15, 2012 19:23, Reimar Döffinger wrote:
> On Tue, May 15, 2012 at 09:27:30AM +0200, Raimund Steger wrote:
>> Apart from the usual glitches on Solaris (need
>> --extra-ldflags="-lX11 -lrt", need --target=sparc-sunos on sun4u,
>
> Interested in trying to get these fixed?

Sure :-)


> I presume the ldflags are only necessary to get X11 support, it
> will compile without?
> And in that case, shouldn't --extra-ldflags="-lrt" work just as well?

I've found that the additional '-lX11' is necessary to compile the gl vo
if the x11 vo isn't enabled. This configuration is the default (without
any special configure arguments) on my Solaris 10 machines. I never
bothered investigating this further, but I've done so now, and I can get
x11 to compile (in which case the extra ldflag isn't necessary) if I

(1) use 'statement_check_broken X11/Xlib.h'... instead of
    'statement_check'... in ./configure, line 4270 (in the
    version from the 1.1 tarball)

(2) add a declaration like:

    extern Status XShmGetEventBase(Display *display);

    to libvo/vo_x11.c, around l. 42.

(Solaris 11 Express and up shouldn't need any of this)


> No idea why the --target is needed though.

If I don't specify it, configure bails out with:

----8<------------

wlm2s092:MPlayer-1.1>./configure \
    --cc=gcc
Checking for gcc version ... 4.5.2
Checking for working compiler ... yes
Detected operating system: SunOS
Detected host architecture: UNKNOWN
Checking for cross compilation ... no
Checking for host cc ... gcc
The architecture of your CPU (UNKNOWN) is not supported by this configure
script
It seems nobody has ported MPlayer to your OS or CPU type yet.

Error: unsupported architecture UNKNOWN

Check "config.log" if you do not understand why it failed.

----8<------------

I think this is because ./configure uses `uname -m` to detect the
architecture, but for Solaris on UltraSPARC this returns 'sun4u', not
'sparc' or 'sparc64' as `uname -p` would.

So one option could probably be to test for 'sun4*', or to use `uname -p`
but this is a bit flakey on Linux, so uname -m is still better.

If there is a need to distinguish the 32-bit from the 64-bit SPARC types
on Solaris (beyond the memory model used for compilation) one would need
to check for something similar than ./configure, l. 2233 already does, i.
e. sun4m|sun4c... vs. sun4u|sun4v.

(I suspect the way this is written at the moment is because Linux and *BSD
also run on SPARC, and I'm not sure what `uname -m` reports for those.)


I'm not sure if this stuff should delay the 1.1 release though.

Raimund


-- 
Worringer Str 31 Duesseldorf 40211 Germany +49-179-2981632 icq 16845346