Re: RFA: GCC 4.2.1: Stabalizing coalesce_list's qsort

[Nick Clifton <nickc at redhat dot com>]

Hi Andrew,

> There is no question that if the sort in unstable, we could end up
> coalescing pairs in different orders, which could then affect the code
> generated. Does this currently affect your bootstrap?

No.  The bug was only exhibited compiling a customer supplied test program with 
a mips64vrel-elf cross compiler.  The compiled program ran if the cross 
compiler had been run under Linux, but failed if the cross compiler was run 
under Cygwin.  This is what lead me to track down the stable qsort issue.

To be honest though I have not tried bootstrapping the 4.2 branch under Cygwin, 
so there is a small chance that the bug might show up there.  I would have 
thought however that at least one Cygwin user would have tried bootstrapping 
4.2 by now and reported an issue if there was a problem.


> Do you have an actual bug or bootstrap failure, or is this meant as a
> preventative measure?   My opinion would be that if we can demonstrate
> it affects bootstrap then its a regression.  Otherwise, any bugs
> uncovered are just that, legitimate bugs that need to be fixed and we
> should fix those rather than cover them up and slow down the sort.  
>
> I'm open for convincing otherwise, but this code has been in there since
> SSA went in for 4.0 I think, and since 4.1 for sure. No problem has been
> previously reported.

That's fine with me.  I suspected that the patch might not be suitable for the 
branch, so I am not worried if it does not go in.  We will maintain a local 
patch to the 4.2 branch sources for our customers and leave it at that.

Cheers
  Nick