You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by rb...@covalent.net on 2001/06/27 22:55:06 UTC
Tagging tree tonight.
Just a heads up. The tree looks to be very stable right now. As soon as
I get home from work today (about 6:00 PST), I plan to commit the threaded
MPM change, test the build one final time, and tag the tree.
If anybody objects let me know. Otherwise, we should have beta candidate
tonight.
Ryan
_____________________________________________________________________________
Ryan Bloom rbb@apache.org
Covalent Technologies rbb@covalent.net
-----------------------------------------------------------------------------
Re: Tagging tree tonight.
Posted by "William A. Rowe, Jr." <ad...@rowe-clan.net>.
This all comes back to the distinction between the max size of the file
to back the mmap, versus the actual size of the mmap'ed view.
This API needs to be totally revisited, but the spirit of setting offhi to zero
works for today.
Bill
----- Original Message -----
From: "William A. Rowe, Jr." <ad...@rowe-clan.net>
To: <ne...@apache.org>
Sent: Wednesday, June 27, 2001 4:34 PM
Subject: Re: Tagging tree tonight.
> Huh? No, DWORD's are 32 bits. apr_off_t is 64 bits. I'll take a look at this
> in the next hour so we can keep going twords a roll.
>
>
> ----- Original Message -----
> From: "Greg Marr" <gr...@alum.wpi.edu>
> To: <ne...@apache.org>
> Sent: Wednesday, June 27, 2001 4:14 PM
> Subject: Re: Tagging tree tonight.
>
>
> > At 05:08 PM 06/27/2001, Bill Stoddard wrote:
> > >Hold off. There is a serious bug in the APR MMAP code on Windows
> > >that is breaking things.
> > >The offending code from mmap.c is here:
> > >
> > > offhi = (DWORD)((*new)->psize << 32);
> >
> > This should be >> instead of <<.
> >
> > #define HIWORD(l) ((WORD)((DWORD_PTR)(l) >> 16))
> >
> > so HIDWORD(l) would be l >> 32
> >
> > --
> > Greg Marr
> > gregm@alum.wpi.edu
> > "We thought you were dead."
> > "I was, but I'm better now." - Sheridan, "The Summoning"
> >
> >
>
>
Re: Tagging tree tonight.
Posted by Greg Marr <gr...@alum.wpi.edu>.
At 05:34 PM 06/27/2001, William A. Rowe, Jr. wrote:
> > At 05:08 PM 06/27/2001, Bill Stoddard wrote:
> > >Hold off. There is a serious bug in the APR MMAP code on Windows
> > >that is breaking things.
> > >The offending code from mmap.c is here:
> > >
> > > offhi = (DWORD)((*new)->psize << 32);
> >
> > This should be >> instead of <<.
> >
> > #define HIWORD(l) ((WORD)((DWORD_PTR)(l) >> 16))
> >
> > so HIDWORD(l) would be l >> 32
>
>DWORD's are 32 bits. apr_off_t is 64 bits.
Right, so when you want the top 32 bits of the 64 bit value, you
shift right 32 bits. (The HIWORD quote was just an analogy).
--
Greg Marr
gregm@alum.wpi.edu
"We thought you were dead."
"I was, but I'm better now." - Sheridan, "The Summoning"
Re: Tagging tree tonight.
Posted by "William A. Rowe, Jr." <ad...@rowe-clan.net>.
Huh? No, DWORD's are 32 bits. apr_off_t is 64 bits. I'll take a look at this
in the next hour so we can keep going twords a roll.
----- Original Message -----
From: "Greg Marr" <gr...@alum.wpi.edu>
To: <ne...@apache.org>
Sent: Wednesday, June 27, 2001 4:14 PM
Subject: Re: Tagging tree tonight.
> At 05:08 PM 06/27/2001, Bill Stoddard wrote:
> >Hold off. There is a serious bug in the APR MMAP code on Windows
> >that is breaking things.
> >The offending code from mmap.c is here:
> >
> > offhi = (DWORD)((*new)->psize << 32);
>
> This should be >> instead of <<.
>
> #define HIWORD(l) ((WORD)((DWORD_PTR)(l) >> 16))
>
> so HIDWORD(l) would be l >> 32
>
> --
> Greg Marr
> gregm@alum.wpi.edu
> "We thought you were dead."
> "I was, but I'm better now." - Sheridan, "The Summoning"
>
>
Re: Tagging tree tonight.
Posted by Greg Marr <gr...@alum.wpi.edu>.
At 05:08 PM 06/27/2001, Bill Stoddard wrote:
>Hold off. There is a serious bug in the APR MMAP code on Windows
>that is breaking things.
>The offending code from mmap.c is here:
>
> offhi = (DWORD)((*new)->psize << 32);
This should be >> instead of <<.
#define HIWORD(l) ((WORD)((DWORD_PTR)(l) >> 16))
so HIDWORD(l) would be l >> 32
--
Greg Marr
gregm@alum.wpi.edu
"We thought you were dead."
"I was, but I'm better now." - Sheridan, "The Summoning"
Re: Tagging tree tonight.
Posted by rb...@covalent.net.
If nothing else happens with this, I will just set to zero before I tag
tonight.
Ryan
On Wed, 27 Jun 2001, Bill Stoddard wrote:
> Hold off. There is a serious bug in the APR MMAP code on Windows that is breaking things.
> The offending code from mmap.c is here:
>
> offhi = (DWORD)((*new)->psize << 32);
>
> psize is a 32 bit integer and shifting it left or right 32 bits apparantly does not change
> the value. So for a 500 byte file, offlo and offhi are both set to 500. Attempting to
> MMAP with the resulting offlo and offhi fails with an insufficient memory.
>
> Unless Mr. Rowe wants to fix this the right way for large file support, I vote for setting
> offhi = 0 then tag and roll.
>
> Bill
>
> ----- Original Message -----
> From: <rb...@covalent.net>
> To: <ne...@apache.org>
> Sent: Wednesday, June 27, 2001 4:55 PM
> Subject: Tagging tree tonight.
>
>
> >
> > Just a heads up. The tree looks to be very stable right now. As soon as
> > I get home from work today (about 6:00 PST), I plan to commit the threaded
> > MPM change, test the build one final time, and tag the tree.
> >
> > If anybody objects let me know. Otherwise, we should have beta candidate
> > tonight.
> >
> > Ryan
> > _____________________________________________________________________________
> > Ryan Bloom rbb@apache.org
> > Covalent Technologies rbb@covalent.net
> > -----------------------------------------------------------------------------
> >
>
>
_____________________________________________________________________________
Ryan Bloom rbb@apache.org
Covalent Technologies rbb@covalent.net
-----------------------------------------------------------------------------
Re: Tagging tree tonight.
Posted by Bill Stoddard <bi...@wstoddard.com>.
Hold off. There is a serious bug in the APR MMAP code on Windows that is breaking things.
The offending code from mmap.c is here:
offhi = (DWORD)((*new)->psize << 32);
psize is a 32 bit integer and shifting it left or right 32 bits apparantly does not change
the value. So for a 500 byte file, offlo and offhi are both set to 500. Attempting to
MMAP with the resulting offlo and offhi fails with an insufficient memory.
Unless Mr. Rowe wants to fix this the right way for large file support, I vote for setting
offhi = 0 then tag and roll.
Bill
----- Original Message -----
From: <rb...@covalent.net>
To: <ne...@apache.org>
Sent: Wednesday, June 27, 2001 4:55 PM
Subject: Tagging tree tonight.
>
> Just a heads up. The tree looks to be very stable right now. As soon as
> I get home from work today (about 6:00 PST), I plan to commit the threaded
> MPM change, test the build one final time, and tag the tree.
>
> If anybody objects let me know. Otherwise, we should have beta candidate
> tonight.
>
> Ryan
> _____________________________________________________________________________
> Ryan Bloom rbb@apache.org
> Covalent Technologies rbb@covalent.net
> -----------------------------------------------------------------------------
>
Re: Tagging tree tonight.
Posted by "Paul J. Reder" <re...@raleigh.ibm.com>.
Good heavens, I set up to test Ryans patch in a heavy load long run
environment. I stopped watching mail at 5:30 and check back this morning
and before any heavy testing could occur everything is committed and tagged.
Whew, that was fast.
I am doing some heavy testing of the current threaded mpm (with Ryan's patch).
I am also approaching completion of my version (still needs testing). I'll
let you all know what I find with the heavy load testing.
--
Paul J. Reder
-----------------------------------------------------------
"The strength of the Constitution lies entirely in the determination of each
citizen to defend it. Only if every single citizen feels duty bound to do
his share in this defense are the constitutional rights secure."
-- Albert Einstein
Re: Tagging tree tonight.
Posted by Bill Stoddard <bi...@wstoddard.com>.
Cliff,
I'll take a look at it a bit more closely when I can. Got a bunch of family obligations
tonight and slipping to the ThinkPad as time allows :-)
Bill
----- Original Message -----
From: "Cliff Woolley" <jw...@virginia.edu>
To: <ne...@apache.org>
Cc: "William Rowe" <wr...@rowe-clan.net>; "Bill Stoddard" <bi...@wstoddard.com>
Sent: Wednesday, June 27, 2001 5:02 PM
Subject: Re: Tagging tree tonight.
> On Wed, 27 Jun 2001 rbb@covalent.net wrote:
>
> > Just a heads up. The tree looks to be very stable right now. As soon as
> > I get home from work today (about 6:00 PST), I plan to commit the threaded
> > MPM change, test the build one final time, and tag the tree.
> >
> > If anybody objects let me know. Otherwise, we should have beta candidate
> > tonight.
>
> +1.
>
> I'd appreciate it if somebody on Win32 could quickly double-check the
> changes I just committed (hint, hint, Bills). They should be fine, but
> verification would be good. The changes assume that the Win32 APR XTHREAD
> files implementation works... is that a safe assumption? If not, we can
> just change the #define in apr.hw for now until it's worked out, and the
> buckets code will handle the XTHREAD files on its own in the meanwhile.
>
> Thanks,
> Cliff
>
>
> --------------------------------------------------------------
> Cliff Woolley
> cliffwoolley@yahoo.com
> Charlottesville, VA
>
>
Re: Tagging tree tonight.
Posted by Cliff Woolley <jw...@virginia.edu>.
On Wed, 27 Jun 2001 rbb@covalent.net wrote:
> Just a heads up. The tree looks to be very stable right now. As soon as
> I get home from work today (about 6:00 PST), I plan to commit the threaded
> MPM change, test the build one final time, and tag the tree.
>
> If anybody objects let me know. Otherwise, we should have beta candidate
> tonight.
+1.
I'd appreciate it if somebody on Win32 could quickly double-check the
changes I just committed (hint, hint, Bills). They should be fine, but
verification would be good. The changes assume that the Win32 APR XTHREAD
files implementation works... is that a safe assumption? If not, we can
just change the #define in apr.hw for now until it's worked out, and the
buckets code will handle the XTHREAD files on its own in the meanwhile.
Thanks,
Cliff
--------------------------------------------------------------
Cliff Woolley
cliffwoolley@yahoo.com
Charlottesville, VA