Mailing List Archive

3 Hiccups with current (Sept. 9) CVS
Isaac (et al.)

Couple of strange errors when compiling current CVS snapshot.


1. If I edit the configure script to include lame support, then 'make', I get:

gcc -c -pipe -w -O3 -DHAVE_AV_CONFIG_H -I.. -D_FILE_OFFSET_BITS=64
-D_LARGEFILE_SOURCE -D_GNU_SOURCE -D_REENTRANT
-D_GNU_SOURCE -DMMX -DQT_NO_DEBUG -DQT_THREAD_SUPPORT
-I../../../../../include -I/usr/local/src/qt/include
-I/usr/local/src/qt/mkspecs/default -o mp3lameaudio.o mp3lameaudio.c
In file included from mp3lameaudio.c:22:
../../../../../include/lame/lame.h:89: parse error before `1'
make[1]: *** [mp3lameaudio.o] Error 1
make[1]: Leaving directory `/usr/local/src/mythCVS/cvs/MC/libavcodec'
make: *** [sub-libavcodec] Error 2

I'm using the most recent version of lame (3.92), and the line in question
from lame.h is:

/*asm optimizations*/
typedef enum asm_optimizations_e {
here-> MMX = 1,
AMD_3DNOW = 2,
SSE = 3
} asm_optimizations;

On the assumption that an enum for MMX and a -DMMX are somehow (??!!??)
confusing each other, if I rip out the -DMMX for just that compile line then
everything is fine. This is a bit odd ....


2. Further along, when building the (new) epg, I get a really weird one:

make[1]: Entering directory `/usr/local/src/mythCVS/cvs/MC/mythepg'
g++ -c -pipe -Wall -W -O6 -march=pentiumpro -fomit-frame-pointer
-funroll-loops -fexpensive-optimizations -finline-functions -D_REENTRANT
-D_GNU_SOURCE -DMMX -DQT_NO_DEBUG -DQT_THREAD_SUPPORT
-I../../../../../include -I../libmythtv -I../libNuppelVideo
-I/usr/local/src/qt/include -I/usr/local/src/qt/mkspecs/default -o
guidegrid.o guidegrid.cpp
/usr/local/src/qt/include/qvaluevector.h: In method
`QValueVectorPrivate<ChannelInfo>::QValueVectorPrivate(const
QValueVectorPrivate<ChannelInfo> &)':
/usr/local/src/qt/include/qvaluevector.h:476: instantiated from
`QValueVector<ChannelInfo>::detachInternal()'
/usr/local/src/qt/include/qvaluevector.h:459: instantiated from here
/usr/local/src/qt/include/qvaluevector.h:125: Internal compiler error.
/usr/local/src/qt/include/qvaluevector.h:125: Please submit a full bug
report.
/usr/local/src/qt/include/qvaluevector.h:125: See
<URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.
make[1]: *** [guidegrid.o] Error 1
make[1]: Leaving directory `/usr/local/src/mythCVS/cvs/MC/mythepg'
make: *** [sub-mythepg] Error 2

I'm on a Pentium 4, and gcc is 2.95.3. If I take out the -O6 optimization
(wow, does the epg really need to be that optimized :-), then it compiles
just fine.

3. Linking for -lvorbis and -lvorbisenc are not automatically included in the
Makefile for ./mythtv/ and ./mythfrontend/


That's it,

- Thor





----------------------------------------------------------------
Thor Sigvaldason <thor@sigvaldason.com>
For my PGP/GnuPG public key, send an e-mail to thorskey@sigvaldason.com
----------------------------------------------------------------
Re: 3 Hiccups with current (Sept. 9) CVS [ In reply to ]
On Monday 09 September 2002 08:50 pm, Thor Sigvaldason wrote:
> Isaac (et al.)
>
> Couple of strange errors when compiling current CVS snapshot.
>
>
> 1. If I edit the configure script to include lame support, then 'make', I
> get:

I should just take that out of the configure script -- it's just for
libavcodec, and I'm not even using that part of it.

> 2. Further along, when building the (new) epg, I get a really weird one:
>
> I'm on a Pentium 4, and gcc is 2.95.3. If I take out the -O6 optimization
> (wow, does the epg really need to be that optimized :-), then it compiles
> just fine.

Heh, probably a old gcc template bug, does -O3 compile fine?

> 3. Linking for -lvorbis and -lvorbisenc are not automatically included in
> the Makefile for ./mythtv/ and ./mythfrontend/

Ah, enabled the vorbis stuff in libavcodec? Also not used =)

Isaac
Re: 3 Hiccups with current (Sept. 9) CVS [ In reply to ]
On Monday 09 September 2002 09:17 pm, you wrote:

> > 2. Further along, when building the (new) epg, I get a really weird one:
> >
> > I'm on a Pentium 4, and gcc is 2.95.3. If I take out the -O6
> > optimization (wow, does the epg really need to be that optimized :-),
> > then it compiles just fine.
>
> Heh, probably a old gcc template bug, does -O3 compile fine?
>

Both -O2 and -O3 generate the same problem as -O6. Weird.

BTW, the new mythepg works just fine, even with all the accented Spanish
stations that I get in my listings. I can finally stop doing that by hand. :-)

Also, did you get the separate e-mail I sent you about MythWeb v0.1?

BR,

Thor

----------------------------------------------------------------
Thor Sigvaldason <thor@sigvaldason.com>
For my PGP/GnuPG public key, send an e-mail to thorskey@sigvaldason.com
----------------------------------------------------------------
Re: 3 Hiccups with current (Sept. 9) CVS [ In reply to ]
On Monday 09 September 2002 09:38 pm, Thor Sigvaldason wrote:
> Both -O2 and -O3 generate the same problem as -O6. Weird.

That it is. I'm not sure how to add a per-file cflag using qmake, else I'd
just turn off optimization for that file..

> BTW, the new mythepg works just fine, even with all the accented Spanish
> stations that I get in my listings. I can finally stop doing that by hand.
> :-)

Cool -- does the OSD also display the proper characters?

> Also, did you get the separate e-mail I sent you about MythWeb v0.1?

Nope, I didn't. Mind resending to ijr@speakeasy.net? One of the cwru.edu
mail servers seems to take awhile sometimes..

Isaac
Re: 3 Hiccups with current (Sept. 9) CVS [ In reply to ]
On Monday 09 September 2002 10:06 pm, you wrote:
> On Monday 09 September 2002 09:38 pm, Thor Sigvaldason wrote:
> > Both -O2 and -O3 generate the same problem as -O6. Weird.
>
> That it is. I'm not sure how to add a per-file cflag using qmake, else I'd
> just turn off optimization for that file..
>

Well, it is easy to fix by hand for the time being.


>
> Cool -- does the OSD also display the proper characters?
>

Yup.


When mythfilldatabase ends, I get a:

Mutex destroy failure: Device or resource busy

Not sure what that is...


> > Also, did you get the separate e-mail I sent you about MythWeb v0.1?
>
> Nope, I didn't. Mind resending to ijr@speakeasy.net? One of the cwru.edu
> mail servers seems to take awhile sometimes..
>

Sent it a few minutes ago.

----------------------------------------------------------------
Thor Sigvaldason <thor@sigvaldason.com>
For my PGP/GnuPG public key, send an e-mail to thorskey@sigvaldason.com
----------------------------------------------------------------
Re: 3 Hiccups with current (Sept. 9) CVS [ In reply to ]
On Mon, 2002-09-09 at 21:38, Thor Sigvaldason wrote:
> Both -O2 and -O3 generate the same problem as -O6. Weird.

-O3 and -O6 are synonymous; -On is equivalent to -O3 for all n > 3. As
well, the only difference between -O2 and -O3 is that -O3 tries to
inline more functions (which can potentially break things if you're
relying on function pointers, etc).

--
Joe Drew <hoserhead@woot.net> <drew@debian.org>

"This particular group of cats is mostly self-herding." -- Bdale Garbee