On Thu, December 26, 2013 20:47, Jan Ehrhardt wrote: > "Anatol Belski" in php.windows (Thu, 26 Dec 2013 19:37:21 +0100): > >> There was a number of commits to opcache last week, I also merged them >> into str_size_and_int64 branch. So the code base should be comparable. I >> think the best way to handle it yet is using bugs.php.net, especially >> as you mean the mainstream is unstable now as well. So if you can get a >> repro case, or even just a BT, it'd be already the pretense to file a >> ticket. > > The mainstream is not unstable (yet?). My own php55 builds include the > latest commits and when running PHP 5.5.7 (x86, TS) as mod_fcgid on my > laptop I do not get crashes, whereas PHP 5.7.9-dev (x86, TS) as mod_fcgid > produces segfaults. > > Faulting application name: php-cgi.exe, version: 5.7.0.0, time stamp: > 0x52bb9102 > Faulting module name: php5ts.dll, version: 5.7.0.0, time stamp: > 0x52bb925d > Exception code: 0xc0000005 > Fault offset: 0x0000df9f > Faulting process id: 0xcc0 > Faulting application start time: 0x01cf02706e9c945d > Faulting application path: D:\phpdev\php57.x86\php-cgi.exe > Faulting module path: D:\phpdev\php57.x86\php5ts.dll > > > I will try to get a BT with GDB. Probably that wil mean I have to load > PHP 5.7 as mod_php in a VC11 Apache, because GDB does not catch errors > of a php-cgi that is loaded from within Apache. > You can attach to the php-cgi.exe process once it's running. Limiting the pool to just one process will ensure there are no other php-cgi processes started. But what do you mean with gdb, under windows? Btw. there's no clue to use the TS version for CGI. Except you wanna use pthreads, maybe. NTS version is also easier to debug. That missing constant phenomenon might be good some caching error. Still, it's better to get a backtrace and file a ticket, so Dmitry or any other opcache maintainers can take a look. Regards Anatol -- PHP Windows Mailing List (http://www.php.net/) To unsubscribe, visit: http://www.php.net/unsub.php
