On Fri, 27 Sep 2013, Benjamin Kaduk wrote:
On Fri, 27 Sep 2013, Jason Edgecombe wrote:
On 09/27/2013 03:09 PM, Benjamin Kaduk wrote:
In other buildbot news, in gerrit/10291, Chas/I have added the
--enable-pthreaded-bos configure option, which is disabled by default. Is
there a reasonable way to have the buildslaves do a build with it enabled?
I think I could create a dummy "do not submit" commit which enables it by
default, just to get the build logs, but I was wondering if there's a
better way.
I can change the compile commands on the slaves, but it depends on how we
want to proceed. Here are some options:
* change one or more builders to enable the configure option
I might be wrong, but I think that passing an unkonwn argument to configure
causes a build failure. So this isn't really an option yet.
* add new builders with the compile options with or without being
gerrit-triggered
* just keep rebasing a gerrit commit as you've done now.
FYI, a builder is just a set commands that is run on one of more build
slaves. Each builder drives a separate compile on the slave(s). If I
schedule pthreaded and non-pthreaded builders on the same slave, then 2
builds will happen on the same slave and possibly slow things down.
Well, tbozo depends on bozo, so there's not really a need for separate
pthreaded and non-pthreaded builders. Once we know that master has it and
builds everywhere, we probably want to just turn it on always, to get the
most build coverage.
Given that, it seems that sticking with "just keep rebasing a gerrit commit"
is the best option for now.
Actually, there is no problem here -- the configure argument just controls
whether we install the threaded versions of things; the makefile logic
tries to build it unconditionally. (Looking at buildbot failures made
clear that this is what was happening.)
Sorry for the noise.
-Ben
_______________________________________________
OpenAFS-devel mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-devel
