Stas Bekman wrote:
Stas Bekman wrote:
Also this core files scan, is it relevant at all to WIN32? When you get a coredump, do you get a core file?
No, there's no core files on Win32. "Dr Watson's log files" would be the closest thing I suppose, but you don't really want to go there. Perhaps if we skip the coredump scan on Win32 then the problem of the "wrong" PID file being picked up would go away?
I suggest we first fix the problem of the wrong path, since it may bite us elsewhere later.
Next please submit a patch to skip core_scan (there are two versions in TestRun), which will make SMOKE much faster on win32.
Argh, looks like I should have waited for the wrong path issue fix, before committing the skip of the core_scan on win32. Applying a reversed patch of this should give you the previous state back:
http://marc.theaimsgroup.com/?l=apache-test-cvs&m=106669945815816&w=2
"Argh", indeed!
I just spent quite a while trying to reproduce & debug the problem with the wrong path, after reversing the patch above, and couldn't for the life of me make it happen any more.
Going back to an earlier version, I see now that you've made some other changes to TestRun to add some "no_chdir => 1" args to the finddepth() calls. Presumably the problem was, as you'd previously suggested, that I had Ctrl+C'ed while a finddepth() was in progress and the current directory had been changed. Stopping the chdir's during the finddepth() call seems to have fixed it, unless I've just been incredibly lucky dozens of times on the trot.
(I wish I'd read the Changes file more carefully -- you did mention the no chdir change, but only as a "speedup", so I didn't take it in properly!)
Cool, so I have fixed the problem as a side-effect of the speedup ;)
But really, that's not the case. In addition to using no_chdir, I'm not not using File::Find in the score can but only check the t/ directory without descending into sub-dirs. But may revert this later. The core file generated by httpd will be dropped only in t/ since that's the server root. Though if any of the .t tests dumping core and happen to chdir out of t/ the core might end up in a different location. Time will show.
So all in all, the current CVS (with the above patch in it) works a treat for me.
I think we've got there at last!
Great!
I'm still very unhappy with the way Ctrl-C works (it wrecks havoc on SMOKE running on unix), but I'll tend to it some time later. At least we have a working version for now.
__________________________________________________________________ Stas Bekman JAm_pH ------> Just Another mod_perl Hacker http://stason.org/ mod_perl Guide ---> http://perl.apache.org mailto:[EMAIL PROTECTED] http://use.perl.org http://apacheweek.com http://modperlbook.org http://apache.org http://ticketmaster.com
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
