Apache OpenOffice (AOO) Bugzilla – Issue 61940
Enable the alternative use of spawn instead of fork/exec for dmake
Last modified: 2013-08-07 15:34:52 UTC
Especially for cygwin it is said that fork/exec is rather slow, but the use of {posix_,}spawn might be also a win for other systems. The following patch adds a new configure option: --enable-spawn (disabled by default) and the correspondend code to dmake to use spawn. The patch seems to be a small win for W32-tcsh, my build time went down from 8h41m to 8h15m, a ~5% speed win. I verified that the patch also works for dmake on linux using the dmake testsuite.
Created attachment 34077 [details] Patch for dmake
Patch committed to dmake43p01, but issue not added due to issue 61940.
Sorry, I meant issue not added due to issue 61941.
Issue added to CWS dmake43p01.
CWS dmake43p01 seems to be growing and growing. Do you have plans for QA and integration?
Hi Stefan, ------- Additional comments from st Wed Feb 15 01:47:07 -0800 2006 ------- CWS dmake43p01 seems to be growing and growing. Do you have plans for QA and integration? yes, was done and integration was planned (from me) to be last year in summer. But with the constrains from RE and the target of OOo 3.0 for the CWS its impossible to integrate it right now. That's the reason why I don't even bother to change the state from "ready for QA" for that CWS. I just continue fixing bugs when I find them. Actually I think there is no reason to block this CWS. The reason was that some functional changes would be introduced by the bugfixing. I didn't know then that this already happened when dmake 4.3 got integrated in cws_src680_ooo20040704. Since then dmake understands: #i17992# Allows the nesting of expressions and a much more complex syntax for dmake. It is now possible to nest expressions with () and any combination of || and &&. But Hamburg's dmake 4.1 doesn't know them. Debian uses dmake from CWS dmake43p01 for it's builds and I certainly also use it for W32-tcsh and W32-4nt builds. There are a few essential patches to fix parallel building it this CWS, without them the build might fail in strange ways. IMHO we should integrate it soon and not wait for later, it cannot hurt Hamburg builds as it's not used there. If there is a consensus to integrate it soon I'll set the CWS to new, finish the remaining issues and then we can do another "build"-QA.
The patch was already committed a while ago to dmake43p01. Reassigning for verification.
@ause: Please verify!
.