You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Branko Čibej <br...@xbc.nu> on 2002/08/27 21:15:57 UTC

Re: svn commit: rev 3075 - trunk trunk/build

> 
>
>+Version 0.14.3 [Alpha Interim 3] (released ?? ??? ????, revision ????)
>+
>+ Developer-visible changes:
>+ * upgrade to neon 0.23.2 to fix Windows build issues
>+
>

I don't know about you, but this commit does _not_ fix the Windows build 
issues. For one thing, build_neon.bat has to be upgraded to set 
EXPAT_FLAGS correcty. But that's trivial -- what's worse is that 
neon-0.23.2 doesn't even compile on Windows when DAV is enabled:

src\ne_socket.c(363) : error C2065: 'EINTR' : undeclared identifier
src\ne_socket.c(410) : error C2065: 'EPIPE' : undeclared identifier


Please revert this change. We're not ready to upgrade Neon yet.

-- 
Brane Čibej   <br...@xbc.nu>   http://www.xbc.nu/brane/


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: rev 3075 - trunk trunk/build

Posted by Justin Erenkrantz <je...@apache.org>.
On Thu, Aug 29, 2002 at 02:05:48PM +0100, Philip Martin wrote:
> Could it be process creation overhead?  How good is Python at starting
> processes on Win32?  Runing the testsuite creates thousands of svn
> client processes.

Processes creation is mighty expensive in Win32.  This is why
httpd has always used threading on Win32 instead of the prefork
model used on most other platforms.  -- justin

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: rev 3075 - trunk trunk/build

Posted by Branko Čibej <br...@xbc.nu>.
Philip Martin wrote:

>Ben Collins-Sussman <su...@collab.net> writes:
>
>  
>
>>brane@xbc.nu writes:
>>
>>    
>>
>>>>I've got a Dell 340 running at 2.0 Ghz and it takes 48 minutes to run with
>>>>the debug build and 45 minutes to run with the release build.
>>>>        
>>>>
>>>I have a 450 MHz PIII, with a terribly slow disk. Around 65 minutes for
>>>ra_local, and 120 minutes (two hours, ouch!) for ra_dav.
>>>      
>>>
>>What a nightmare!  On FreeBSD it only takes me about 20 mins, and
>>about 10-15 mins for Mike and Karl on Linux.  (These are all 700mhz
>>machines with ordinary ide disks.)  Is there *that* big a difference
>>in I/O between os'es?  Maybe APR's I/O on win32 is really badly
>>implemented?
>>    
>>
>
>Could it be process creation overhead?  How good is Python at starting
>processes on Win32?  Runing the testsuite creates thousands of svn
>client processes.
>

It's probably both. NT is notoriously slow about process creation 
(because there's so much crud hung off a Windows process), and while 
file I/O is quite fast, creating and opening files isn't. So ... i need 
a sponsor for a 2-CPU box with hardware RAID ...

-- 
Brane Čibej   <br...@xbc.nu>   http://www.xbc.nu/brane/


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: rev 3075 - trunk trunk/build

Posted by Philip Martin <ph...@codematters.co.uk>.
Ben Collins-Sussman <su...@collab.net> writes:

> brane@xbc.nu writes:
> 
> > > I've got a Dell 340 running at 2.0 Ghz and it takes 48 minutes to run with
> > > the debug build and 45 minutes to run with the release build.
> > 
> > I have a 450 MHz PIII, with a terribly slow disk. Around 65 minutes for
> > ra_local, and 120 minutes (two hours, ouch!) for ra_dav.
> 
> What a nightmare!  On FreeBSD it only takes me about 20 mins, and
> about 10-15 mins for Mike and Karl on Linux.  (These are all 700mhz
> machines with ordinary ide disks.)  Is there *that* big a difference
> in I/O between os'es?  Maybe APR's I/O on win32 is really badly
> implemented?

Could it be process creation overhead?  How good is Python at starting
processes on Win32?  Runing the testsuite creates thousands of svn
client processes.

-- 
Philip Martin

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: rev 3075 - trunk trunk/build

Posted by Ben Collins-Sussman <su...@collab.net>.
brane@xbc.nu writes:

> > I've got a Dell 340 running at 2.0 Ghz and it takes 48 minutes to run with
> > the debug build and 45 minutes to run with the release build.
> 
> I have a 450 MHz PIII, with a terribly slow disk. Around 65 minutes for
> ra_local, and 120 minutes (two hours, ouch!) for ra_dav.

What a nightmare!  On FreeBSD it only takes me about 20 mins, and
about 10-15 mins for Mike and Karl on Linux.  (These are all 700mhz
machines with ordinary ide disks.)  Is there *that* big a difference
in I/O between os'es?  Maybe APR's I/O on win32 is really badly
implemented?


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: rev 3075 - trunk trunk/build

Posted by br...@xbc.nu.
Quoting Blair Zajac <bl...@orcaware.com>:

> brane@xbc.nu wrote:
> > Out of interest, how long does your run of win-tests.py take? (and on
> > what kind of machine).
> 
> I've got a Dell 340 running at 2.0 Ghz and it takes 48 minutes to run with
> the debug build and 45 minutes to run with the release build.

Ouch! I think we're horribly I/O bound on those tests.
 
> How long on your box?

I have a 450 MHz PIII, with a terribly slow disk. Around 65 minutes for
ra_local, and 120 minutes (two hours, ouch!) for ra_dav.

    Brane

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: rev 3075 - trunk trunk/build

Posted by Blair Zajac <bl...@orcaware.com>.
brane@xbc.nu wrote:
> 
> > Oh trust me, I test all my patches.  Waiting around for win-tests.py to
> > finish is painful, but well worth it.
> 
> Out of interest, how long does your run of win-tests.py take? (and on what kind
> of machine).

I've got a Dell 340 running at 2.0 Ghz and it takes 48 minutes to run with
the debug build and 45 minutes to run with the release build.

How long on your box?

> 
> > Also, we're using different build platform from each other, so unless
> > either one of us gets both Visual Studio.net and Visual Studio 6 on the
> > same system, we'll run into this every so often.
> 
> I hope it's every so *seldom*. :-)

Me too.

Best,
Blair

-- 
Blair Zajac <bl...@orcaware.com>
Web and OS performance plots - http://www.orcaware.com/orca/

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: rev 3075 - trunk trunk/build

Posted by br...@xbc.nu.
Quoting Blair Zajac <bl...@orcaware.com>:

> > And I'll continue to do so every time somebody breaks the build for no
> > obvious reason (obvious to me, that is). That'll teach people to test
> > their patches before committing them. ;-)
> 
> And expect the reverse :)

Oh, absolutely. As long as it doesn't degenerate into a veto war like the one
that happened on dev@apr a while ago. :-)


> Oh trust me, I test all my patches.  Waiting around for win-tests.py to
> finish is painful, but well worth it.

Out of interest, how long does your run of win-tests.py take? (and on what kind
of machine).
 
> Also, we're using different build platform from each other, so unless
> either one of us gets both Visual Studio.net and Visual Studio 6 on the
> same system, we'll run into this every so often.

I hope it's every so *seldom*. :-)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: rev 3075 - trunk trunk/build

Posted by Blair Zajac <bl...@orcaware.com>.
Branko ?ibej wrote:
> 
> Blair Zajac wrote:
> 
> >Branko ?ibej wrote:
> >
> >
> >>>+Version 0.14.3 [Alpha Interim 3] (released ?? ??? ????, revision ????)
> >>>+
> >>>+ Developer-visible changes:
> >>>+ * upgrade to neon 0.23.2 to fix Windows build issues
> >>>+
> >>>
> >>>
> >>>
> >>I don't know about you, but this commit does _not_ fix the Windows build
> >>issues. For one thing, build_neon.bat has to be upgraded to set
> >>EXPAT_FLAGS correcty. But that's trivial -- what's worse is that
> >>neon-0.23.2 doesn't even compile on Windows when DAV is enabled:
> >>
> >>src\ne_socket.c(363) : error C2065: 'EINTR' : undeclared identifier
> >>src\ne_socket.c(410) : error C2065: 'EPIPE' : undeclared identifier
> >>
> >>
> >>
> >
> >I've built and compiled svn with Neon just fine, but then I always
> >build with zlib and openssl support.  I just tried to build Neon
> >without openssl and it failed with the same errors you had.
> >
> >Try this patch and let me know how it works.
> >
> 
> Ah, great!. Works for me. Go ahead and commit this, if you will.

Will do.

> 
> >Index: build/win32/build_neon.bat
> >
> [snip]
> 
> >>Please revert this change. We're not ready to upgrade Neon yet.
> >>
> >>
> >
> >The solution to everything is not to revert (this is the second revert
> >you've suggested recently)!
> >
> >
> And I'll continue to do so every time somebody breaks the build for no
> obvious reason (obvious to me, that is). That'll teach people to test
> their patches before committing them. ;-)

And expect the reverse :)

Oh trust me, I test all my patches.  Waiting around for win-tests.py to
finish is painful, but well worth it.

Also, we're using different build platform from each other, so unless
either one of us gets both Visual Studio.net and Visual Studio 6 on the
same system, we'll run into this every so often.

Best,
Blair

-- 
Blair Zajac <bl...@orcaware.com>
Web and OS performance plots - http://www.orcaware.com/orca/

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: rev 3075 - trunk trunk/build

Posted by Branko Čibej <br...@xbc.nu>.
Blair Zajac wrote:

>Branko ?ibej wrote:
>  
>
>>>+Version 0.14.3 [Alpha Interim 3] (released ?? ??? ????, revision ????)
>>>+
>>>+ Developer-visible changes:
>>>+ * upgrade to neon 0.23.2 to fix Windows build issues
>>>+
>>>
>>>      
>>>
>>I don't know about you, but this commit does _not_ fix the Windows build
>>issues. For one thing, build_neon.bat has to be upgraded to set
>>EXPAT_FLAGS correcty. But that's trivial -- what's worse is that
>>neon-0.23.2 doesn't even compile on Windows when DAV is enabled:
>>
>>src\ne_socket.c(363) : error C2065: 'EINTR' : undeclared identifier
>>src\ne_socket.c(410) : error C2065: 'EPIPE' : undeclared identifier
>>
>>    
>>
>
>I've built and compiled svn with Neon just fine, but then I always
>build with zlib and openssl support.  I just tried to build Neon
>without openssl and it failed with the same errors you had.
>
>Try this patch and let me know how it works.
>

Ah, great!. Works for me. Go ahead and commit this, if you will.

>Index: build/win32/build_neon.bat
>
[snip]

>>Please revert this change. We're not ready to upgrade Neon yet.
>>    
>>
>
>The solution to everything is not to revert (this is the second revert
>you've suggested recently)!
>  
>
And I'll continue to do so every time somebody breaks the build for no 
obvious reason (obvious to me, that is). That'll teach people to test 
their patches before committing them. ;-)


-- 
Brane Čibej   <br...@xbc.nu>   http://www.xbc.nu/brane/


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: rev 3075 - trunk trunk/build

Posted by Blair Zajac <bl...@orcaware.com>.
Branko ?ibej wrote:
> 
> >
> >
> >+Version 0.14.3 [Alpha Interim 3] (released ?? ??? ????, revision ????)
> >+
> >+ Developer-visible changes:
> >+ * upgrade to neon 0.23.2 to fix Windows build issues
> >+
> >
> 
> I don't know about you, but this commit does _not_ fix the Windows build
> issues. For one thing, build_neon.bat has to be upgraded to set
> EXPAT_FLAGS correcty. But that's trivial -- what's worse is that
> neon-0.23.2 doesn't even compile on Windows when DAV is enabled:
> 
> src\ne_socket.c(363) : error C2065: 'EINTR' : undeclared identifier
> src\ne_socket.c(410) : error C2065: 'EPIPE' : undeclared identifier
> 

I've built and compiled svn with Neon just fine, but then I always
build with zlib and openssl support.  I just tried to build Neon
without openssl and it failed with the same errors you had.

Try this patch and let me know how it works.

Index: build/win32/build_neon.bat
===================================================================
--- build/win32/build_neon.bat
+++ build/win32/build_neon.bat	2002-08-27 16:36:51.000000000 -0700
@@ -11,8 +11,14 @@
 @rem * NOTE: The paths should be relative to the Neon directory, ..\..\neon
 @rem **************************************************************************
 
-@rem Must set EXPAT_FLAGS, not EXPAT_SRC, to define HAVE_EXPAT_H
-set EXPAT_FLAGS="/I ../apr-util/xml/expat/lib /D HAVE_EXPAT_H"
+@rem Must use EXPAT_FLAGS and not EXPAT_SRC to set the include location
+@rem correctly for APR's expat and define HAVE_EXPAT and HAVE_EXPAT_H,
+@rem since Neon's neon.mak is set up to use the pre-compiled Windows
+@rem binary Expat installation.
+@rem Also define HAVE_ERRNO_H to work around a bug in 0.23.2 when
+@rem compiling without OpenSSL, where EINTR and EPIPE are not defined
+@rem in src/ne_socket.c unless errno.h is included.
+set EXPAT_FLAGS="/I ../apr-util/xml/expat/lib /D HAVE_EXPAT /D HAVE_EXPAT_H /D HAVE_ERRNO_H"
 
 set exitcode=0
 

> Please revert this change. We're not ready to upgrade Neon yet.

The solution to everything is not to revert (this is the second revert
you've suggested recently)!

Best,
Blair

-- 
Blair Zajac <bl...@orcaware.com>
Web and OS performance plots - http://www.orcaware.com/orca/

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org