On Sat, Aug 6, 2011 at 2:12 AM, Soeren Sandmann <[email protected]> wrote: > Andrea Canciani <[email protected]> writes: > >> On Thu, Aug 4, 2011 at 6:32 AM, Soeren Sandmann <[email protected]> wrote: >>> Søren Sandmann <[email protected]> writes: >>> >>>> From: Søren Sandmann Pedersen <[email protected]> >>>> >>>> This test generates random regions and checks whether random boxes and >>>> points are contained within them. The results are combined and a CRC32 >>>> value is computed and compared to a known-correct one. >>> >>> Apparently, I forgot to add the region-contains-test.c file, but it was >>> buggy anyway and had endian issues. Below is a new version of that patch. >> >> In the endianness test, why is the variable marked as volatile? >> (I'm not saying it's wrong, I'm just saying I don't understand it) >> Apart from this and some trailing whitespace, >> >> Reviewed-by: Andrea Canciani <[email protected]> > > The endian test was written by Siarhei originally, but my guess is that > without it, compilers might decide to optimize the whole thing away, > which could be an issue on architectures that support both big and > little endian modes.
I'm a little worried by this, because most of pixman (and cairo) #ifdef's out code depending on the endianness detected by the compiler. Anyway, regarding this patch, the worst case is that this check is more paranoid than needed. I think we might want to investigate if this check is also needed elsewhere. "git am" still complains about trailing whitespace in patch 2/4 version 2: /home/ranma42/Code/fdo/pixman/.git/rebase-apply/patch:128: trailing whitespace. /home/ranma42/Code/fdo/pixman/.git/rebase-apply/patch:163: trailing whitespace. Apart from this, for the series: Reviewed-by: Andrea Canciani <[email protected]> Andrea _______________________________________________ Pixman mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/pixman
