Trying to package/port projects using optimizations implemented in
assembler I encounter the same issue for which I didn't found a solution.
For the moment, the issue manifests itself only on i386 architecture as
the porting has as source, projects developed on Linux and concerns
mainly video applications, e.g., ffmpeg, xvid, &c.
In the linking phase of the share library I get the following message:
Text relocation remains referenced
against symbol offset in file
<unknown> 0x7 bitstream/x86_asm/cbp_mmx.o
.
.
.
<unknown> 0xc6 utils/x86_asm/cpuid.o
ld: fatal: relocations remain against allocatable but non-writable sections
collect2: ld returned 1 exit status
The result of my research can be summarized as follows: don't
use hand written assembly in shared libraries if you're not using the
GNU linker.
The consequences of this are:
- don't activate assembler optimizations: poor performance
- where the previous is not possible, don't generate shared
libraries containing the assembler code: build only statically linked
binaries --- at least the corresponding library is statically linked
Is there a fellow maintainer having a positive experience on building
shared libraries for Solaris containing assembler code and willing to
share his solution?
--
Peter
_______________________________________________
maintainers mailing list
[email protected]
https://lists.opencsw.org/mailman/listinfo/maintainers
