Axel.Thimm at ATrpms
Sep 2, 2009, 11:24 AM
Post #6 of 9
(1246 views)
Permalink
|
On Wed, Sep 02, 2009 at 01:18:45PM -0400, Brian Long wrote:
> On Wed, Sep 2, 2009 at 10:21 AM, Axel Thimm <Axel.Thimm [at] atrpms> wrote:
>
> > On Tue, Sep 01, 2009 at 03:15:06AM -0700, Yeechang Lee wrote:
> > > Axel Thimm <Axel.Thimm [at] ATrpms> says:
> > > > I think mythtv already builds with -fPIC, I'll have to check the
> > > > buildlogs to be sure.
> > >
> > > My system indeed builds the SRPM with -fPIC.
> >
> > I can confirm this - and Brian's mythtv, who is using F11/i586 had
> > this in the buildlog:
> >
> > gcc -c -pipe -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions
> > -fstack-pro
> > tector --param=ssp-buffer-size=4 -m32 -march=i586 -mtune=generic
> > -fasynchronous-
> > unwind-tables -fomit-frame-pointer -pthread -Wall -Wno-switch
> > -Wdisabled-optimiz
> > ation -Wpointer-arith -Wredundant-decls -Wno-pointer-sign
> > -Wno-missing-prototype
> > s -D_REENTRANT -DPIC -fPIC -DMMX -Di386 -D_GNU_SOURCE
> > -D_FILE_OFFSET_BITS=64 -D
> > QT_NO_DEBUG -DQT_THREAD_SUPPORT -DQT_PLUGIN -DQT_SHARED -DQT_TABLET_SUPPORT
> > -I/u
> > sr/lib/qt-3.3/mkspecs/default -I. -I/usr/include -I/usr/include
> > -I/usr/include -
> > I../../libs/libavcodec -I../../libs/libavutil -I../../libs/libmythtv
> > -I../../lib
> > s/libavcodec -I../.. -I/usr/lib/qt-3.3/include -o filter_greedyhdeint.o
> > filter_g
> > reedyhdeint.c
> >
> > So the flags are already in place. Something else seems to be wrong.
> >
>
> Does this help?
> > eu-findtextrel /usr/lib/mythtv/filters/libgreedyhdeint.so
> either the file containing the function 'GreedyHDeintFilter' or the file
> containing the function 'apply_chroma_filter' is not compiled with
> -fpic/-fPIC
>
> I'm not too familiar with -fPIC and -fPIE, but I ran the utility mentioned
> at the redhat.com URL and it output the above line multiple times. It
> appears apply_chroma_filter is contained in color.c. Are the build logs
> publicly available?
No, they are not, and they should, so I don't have to dig them up :)
color.c was built with the same options:
gcc -c -pipe -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-pro
tector --param=ssp-buffer-size=4 -m32 -march=i586 -mtune=generic -fasynchronous-
unwind-tables -fomit-frame-pointer -pthread -Wall -Wno-switch -Wdisabled-optimiz
ation -Wpointer-arith -Wredundant-decls -Wno-pointer-sign -Wno-missing-prototype
s -D_REENTRANT -DPIC -fPIC -DMMX -Di386 -D_GNU_SOURCE -D_FILE_OFFSET_BITS=64 -D
QT_NO_DEBUG -DQT_THREAD_SUPPORT -DQT_PLUGIN -DQT_SHARED -DQT_TABLET_SUPPORT -I/u
sr/lib/qt-3.3/mkspecs/default -I. -I/usr/include -I/usr/include -I/usr/include -
I../../libs/libavcodec -I../../libs/libavutil -I../../libs/libmythtv -I../../lib
s/libavcodec -I../.. -I/usr/lib/qt-3.3/include -o color.o color.c
So anything that libgreedyhdeint.so is built from is built with -fPIC,
but AFAIK if there is some asm in there it can make it non-relocatable
as well. And both color.c and filter_greedyhdeint.c (which includes
greedyh.asm) have asm code inside. If one wants to fix this one needs
to fix upstream the asm code. The Red Hat URL further quotes on that:
The tool is not able to distinguish compiler-generated code
from assembler code written by the programmer. In both cases
the tool will print the message above since it is what is
correct in 99% of all cases. If the problem is indeed the
result of hand-written assembler code the solution is not as
simple as adding a compiler/assembler flag. The code needs to
be rewritten. This is architecture specific and can vary widely
between every single instance. We are not going into those
details here. Find a person with sufficient assembly
programming skills if this problem appears.
--
Axel.Thimm at ATrpms.net
|
