Re: RFC Fix for 9964.cc and 9507.cc on ia64-hpux

[Benjamin Kosnik <bkoz at redhat dot com>]

>Thanks.  I've augmented this into what I hope are complete patches for
>3.4/mainline and 3.3; attached.  I've dropped out the alarm/wait bits,
>including only the open_mode::flags fixes to the testsuite; see below.
>Testing running on ia64-hpux now.

Ok.

For the 3.4 patches, kill all 

        * config/abi/alpha-freebsd5/baseline_symbols.txt
        * config/abi/alpha-linux-gnu/baseline_symbols.txt
        * config/abi/hppa-linux-gnu/baseline_symbols.txt
        * config/abi/i386-freebsd4/baseline_symbols.txt
        * config/abi/i386-freebsd5/baseline_symbols.txt
        * config/abi/i486-linux-gnu/baseline_symbols.txt
        * config/abi/ia64-linux-gnu/baseline_symbols.txt
        * config/abi/mips-linux-gnu/baseline_symbols.txt
        * config/abi/sparc-freebsd5/baseline_symbols.txt
        * config/abi/sparc-linux-gnu/baseline_symbols.txt
        * config/abi/x86_64-linux-gnu/baseline_symbols.txt:
	Remove entry for __basic_file<char>::_M_open_mode.

These have no meaning on the 3.4 branch until there is a 3.4.0 to
establish a 3.4.0 ABI export list, so there is no point in editing them.

Also, minor minor 

+  static const char*
+  fopen_mode(std::ios_base::openmode mode)

can just be

+  const char*
+  fopen_mode(std::ios_base::openmode mode)

The 3.3 patch looks great.

Other than that, everything is fine and can go in, thanks.

>This is nontrivial.  alarm() doesn't inherit across fork() so doing it
>in try_mkfifo isn't good enough.  I can try to come up with a good set
>of wrappers for this but I don't think it should block this patch.

Agreed.

-benjamin