performance enhancements for cygwin make

Noel Grandin noelgrandin at
Mon Mar 12 02:06:12 PDT 2012


OK, I got a version up and running after hacking on the version of gmake 
from the libreoffice repo.

As this point in time, my opinion is that the idea is not worth doing - 
things like touch and cp don't actually seem to trigger very often in 
our build process.
(Unless I'm doing something very wrong).

So I'm going to drop this approach for now.

What I am going to do is run a Windows profiler and see what system 
calls are being triggered by gmake.
I seem to remember seeing something on the GIT mailing list where they 
bypassed the CYGWIN emulation layer to use underlying Windows system 
calls because the default CYGWIN stat() implementation loads a lot more 
information than is actually required for the purposes of make.

-- Noel Grandin

On 2012-03-08 11:07, Michael Meeks wrote:
> On Thu, 2012-03-08 at 09:52 +0200, Noel Grandin wrote:
>> I'm having a bash at this - mostly I copied and simplified the
>> corresponding code from the GNU touch utility.
> 	ooh - exciting ! :-) thanks for that, I'll hold fire on filing the easy
> hack on that. Hopefully you're working on the version from our git repo
> that is at least sanely revision controlled, fast local diffs etc. :-)
>> Anybody have any ideas?
> 	I think the consensus is right - that you should write them for unix as
> well, although the win is smaller there it will be at least something I
> guess.
> 	Looking forward to your patches ! :-)
> 	Thanks !
> 		Michael.

More information about the LibreOffice mailing list