apr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mladen Turk" <mt...@mappingsoft.com>
Subject Re: [PATCH] WIN32 Optimistic apr_proc_kill
Date Sat, 16 Feb 2002 19:28:07 GMT

----- Original Message -----
From: "William A. Rowe, Jr." <wrowe@covalent.net>
To: "Mladen Turk" <mturk@mappingsoft.com>; "Bill Tutt" <rassilon@lyra.org>;
Sent: Saturday, February 16, 2002 7:43 PM
Subject: Re: [PATCH] WIN32 Optimistic apr_proc_kill

> From: "Mladen Turk" <mturk@mappingsoft.com>
> Sent: Saturday, February 16, 2002 12:26 PM
> > I'm sorry if my English Leeds to some misunderstanding.
> So does mine, ask most English speakers :)
Yes but I'm runing the spell checking on each mail I send, and still find
myself lughing!

> MOST apps will not do their own cleanup within DllLoad proc
> The question is first fixing the most common cases (close the doors on
> with stdin/stdout, send Ctrl+Break or WM_EXIT, and pause a moment.)  Then
> investigate the most thorough way to clobber them hard, but thoroughly :)

That's exacty what I'm saying... We could try to send CTRL+WHATEVER but that
wouldn't kill the Excell for example. How the sending WM_EXIT diferentiates
from ExitProcess? You still find yourself inside some "Are You Sure" loop.
We could try to do that, but evetualy will found ousrelf with the unresolved
WM_XXX message.
IMO we still need the kill-proc-tree.
Couple of months ago someone told me that other-child performs that, but if
I kill the apache the rotatelogs will still keep hanging forever.
Now I think that Process32First, Process32Next will solve some of the
problems I have.

> And we cannot cleanup apr this way either; since the clib will be gone
> beneath us before we got that far.

Sorry Bill but you are wrong about that. The APR is called from inside
dllMainCRTStartup... if we are dll, but anyhow the clib gets freed as last,
the catch is in the "atexit"...



View raw message