Ruediger Pluem wrote: > > On 08/06/2007 04:20 PM, [EMAIL PROTECTED] wrote: >> Author: jim >> Date: Mon Aug 6 07:20:24 2007 >> New Revision: 563147 >> >> URL: http://svn.apache.org/viewvc?view=rev&rev=563147 >> Log: >> These atomics expect apr_uint32_t *... The expectation, of course, >> is that the add/inc still works "as expected" even though we >> are using signed values. > > Hm. I have a bad feeling here. These atomics are platform specific and seemed > to be designed for unsigned ints. Are we really sure that they can handle > signed ints > as well? Comments from an atomics guru from APR?
I'm no guru on the subject, but AFAIK it's not reliable because of the underlying signed <-> unsigned conversions and unsigned arithmetic of the various instructions used. Looking at the code, I think we could use some other scheme to track the blocked threads. -- Davi Arnaut
