David is already looking into reversing the command substitution implementation to use files and mmap() if the POSIX standard allows this. That would solve this problem and the performance regressions.
Olga On Mon, May 3, 2010 at 1:19 PM, Siddhesh Poyarekar <[email protected]> wrote: > On Wed, Apr 28, 2010 at 7:29 PM, Siddhesh Poyarekar > <[email protected]> wrote: >> This is most likely an OS level limit. The number comes up different >> for different versions. It is 19600 for my RHEL-5 32 bit test box, >> 22331 for my RHEL-5 64 bit and 21648 for my Fedora 64 bit. Both RHEL-5 >> boxes run 20100202 and the Fedora box runs with a patched version of >> 20100309. >> > > The limit is determined by /proc/sys/net/core/wmem_default in Linux. > The problem is that the parent blocks till the child exits and then > starts reading, so in case the buffer gets filled up, we have a > deadlock. > > > -- > Siddhesh Poyarekar > http://siddhesh.in > _______________________________________________ > ast-developers mailing list > [email protected] > https://mailman.research.att.com/mailman/listinfo/ast-developers > -- , _ _ , { \/`o;====- Olga Kryzhanovska -====;o`\/ } .----'-/`-/ [email protected] \-`\-'----. `'-..-| / Solaris/BSD//C/C++ programmer \ |-..-'` /\/\ /\/\ `--` `--` _______________________________________________ ast-developers mailing list [email protected] https://mailman.research.att.com/mailman/listinfo/ast-developers
