This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: powerpc64 stabs fix.

At 06:32 AM 9/13/2002, Alan Modra wrote:
-gstabs+ on powerpc64-linux causes the assembler to complain, eg.
/tmp/ccRfbYgd.s:133: Error: subtraction of two symbols in different
sections `.text' {.text section} - `main' {.opd section} at file
address 0000000000000230

I missed this case when fixing a similar -gstabs problem.

* doc/tm.texi (DBX_OUTPUT_NFUN): Describe.
* dbxout.c (dbxout_function_end): Use DBX_OUTPUT_NFUN.
* config/rs6000/linux64.h (DBX_OUTPUT_NFUN): Define.

OK to apply?

Incidentally, DBX_OUTPUT_FUNCTION_END seems to be unused. I'll happily
remove it too if requested.
./gcc/config/d30v/d30v.h:/* #define DBX_OUTPUT_FUNCTION_END(STREAM, FUNCTION) */
./gcc/config/stormy16/stormy16.h:/* #define DBX_OUTPUT_FUNCTION_END(STREAM, FUNCTION) */
./gcc/dbxout.c:#ifdef DBX_OUTPUT_FUNCTION_END
./gcc/dbxout.c: DBX_OUTPUT_FUNCTION_END (asmfile, decl);
./gcc/doc/tm.texi:@findex DBX_OUTPUT_FUNCTION_END
./gcc/doc/tm.texi:@item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
Please don't.  I need that for VAX ELF support and stabs.  I have to define
that as function_section (decl).  Otherwise, if the function had static
variables then gcc would emit the function-ending stab in the wrong section.
Now if you changes dbxout_function_end to always call function_decl (decl)
before emitting the stab, then it could go away.  It might be interesting to
see if you did the above, whether your patch is even needed.

Matt Thomas               Internet:
3am Software Foundry      WWW URL:
Cupertino, CA             Disclaimer: I avow all knowledge of this message

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]