|
From: Julian S. <js...@ac...> - 2007-11-22 18:53:49
|
Greetings all. I'm trying to wrap up the long overdue 3.3.0 release. It's been nearly 18 months since 3.2.0 shipped. The 3.2.X line has done well, but it's a bit old now. Plus we've accumulated a bunch of cool new stuff to ship over the past 18 months: * revised Massif (almost a complete rewrite) * Helgrind works again (a complete rewrite) * Cachegrind does branch-misprediction profiling * experimental tools section (Omega, DRD) * modestly restructured/rationalised documentation * support for latest toolchains/libs: gcc-4.3, glibc-2.7 * many bugs fixed, as usual * scalability improvements (for v. large programs); some modest performance improvements I would like to have a release candidate ready by Fri 30 Nov, with a final release a week later, on Fri 7 December. Naturally, if there is any critical breakage that should be fixed before 3.3.0, please shout now. J |
|
From: Florian K. <br...@ac...> - 2007-11-23 05:22:11
|
On Thursday 22 November 2007 1:53 pm, Julian Seward wrote: > Greetings all. > > I'm trying to wrap up the long overdue 3.3.0 release. It's been > nearly 18 months since 3.2.0 shipped. The 3.2.X line has done well, > but it's a bit old now. Plus we've accumulated a bunch of cool new > stuff to ship over the past 18 months: > > * revised Massif (almost a complete rewrite) > * Helgrind works again (a complete rewrite) > * Cachegrind does branch-misprediction profiling > * experimental tools section (Omega, DRD) Do you have plans to morph branches/ORIGIN_TRACKING into an experimental tool? I read the pldi paper and understand it has limitations but it could still make a quite a difference if a bug is of the kind where it can track its origin. I've checked out the branch to play with it a bit, but unfortunately it does not compile, due to changes in VEX IR. Florian |
|
From: Nicholas N. <nj...@cs...> - 2007-11-23 08:52:02
|
On Thu, 22 Nov 2007, Florian Krohm wrote: > Do you have plans to morph branches/ORIGIN_TRACKING into an > experimental tool? I read the pldi paper and understand it has > limitations but it could still make a quite a difference if a > bug is of the kind where it can track its origin. > I've checked out the branch to play with it a bit, but > unfortunately it does not compile, due to changes in VEX IR. No... the cost/benefit trade-off wasn't very good -- it added quite a bit of complication to Memcheck and the core. N |
|
From: Bart V. A. <bar...@gm...> - 2007-11-30 18:33:32
|
On Nov 22, 2007 7:53 PM, Julian Seward <js...@ac...> wrote: > > Greetings all. > > I'm trying to wrap up the long overdue 3.3.0 release. Regarding the 3.3.0 release, a question about the nightly build: from the nightly build output I can see that the exp-drd source code and regression tests are compiled, but are the exp-drd regression tests run during the nightly build ? I could not find any reference to exp-drd's regression test results in the last few nightly build e-mails. The nightly build mails I receive indicate that the nightly build is run on the following systems: * alvis ( i686, Red Hat 7.3 ) * lloyd ( x86_64, Fedora 7 ) * dellow ( x86_64, Fedora 8 ) * gill ( x86_64, Fedora Core 2 ) * g5 ( SuSE 10.1, ppc970 ) Regards, Bart. |
|
From: Nicholas N. <nj...@cs...> - 2007-12-03 08:35:58
|
On Fri, 30 Nov 2007, Bart Van Assche wrote: > Regarding the 3.3.0 release, a question about the nightly build: from > the nightly build output I can see that the exp-drd source code and > regression tests are compiled, but are the exp-drd regression tests > run during the nightly build ? I could not find any reference to > exp-drd's regression test results in the last few nightly build > e-mails. > > The nightly build mails I receive indicate that the nightly build is > run on the following systems: > * alvis ( i686, Red Hat 7.3 ) > * lloyd ( x86_64, Fedora 7 ) > * dellow ( x86_64, Fedora 8 ) > * gill ( x86_64, Fedora Core 2 ) > * g5 ( SuSE 10.1, ppc970 ) It seems the regtests are split in two -- controlled by 'make regtest' for the normal tools and 'make exp-regtest' for the experimental tools. The nightly script doesn't run 'make exp-regtest'. I can either change the script to run 'make exp-regtest' -- which will require Tom and Julian to update their copy of the script on those machines -- or we can change 'make regtest' run all of them. I'm leaning towards the latter, as an incentive to get the exp-tools' tests working as well as possible. Julian, what do you think? N |
|
From: Tom H. <to...@co...> - 2007-12-03 08:47:51
|
In message <Pin...@mu...>
Nicholas Nethercote <nj...@cs...> wrote:
> I can either change the script to run 'make exp-regtest' -- which will
> require Tom and Julian to update their copy of the script on those machines
> -- or we can change 'make regtest' run all of them. I'm leaning towards the
> latter, as an incentive to get the exp-tools' tests working as well as
> possible. Julian, what do you think?
My machines actually check the script out of the repository each
night, so just updating the repository to run both would fix my
machines.
Tom
--
Tom Hughes (to...@co...)
http://www.compton.nu/
|
|
From: Julian S. <js...@ac...> - 2007-12-03 12:35:59
|
On Monday 03 December 2007 09:47, Tom Hughes wrote: > In message <Pin...@mu...> > > Nicholas Nethercote <nj...@cs...> wrote: > > I can either change the script to run 'make exp-regtest' -- which will > > require Tom and Julian to update their copy of the script on those > > machines -- or we can change 'make regtest' run all of them. I'm leaning > > towards the latter, as an incentive to get the exp-tools' tests working > > as well as possible. Julian, what do you think? I'd say, for now change the script to run make exp-regtest too. Tom's machines will auto-update and I'll change mine to do whatever Tom's are doing. After 3.3.0 is done we can merge make exp-regtest into standard make regtest. > My machines actually check the script out of the repository each > night, so just updating the repository to run both would fix my > machines. That's neat. So the top level cron command runs some other little script which does the checkout? Can you send the script? J |
|
From: Bart V. A. <bar...@gm...> - 2008-01-07 20:52:31
|
On Dec 3, 2007 1:35 PM, Julian Seward <js...@ac...> wrote: > On Monday 03 December 2007 09:47, Tom Hughes wrote: > > In message <Pin...@mu...> > > > > Nicholas Nethercote <nj...@cs...> wrote: > > > I can either change the script to run 'make exp-regtest' -- which will > > > require Tom and Julian to update their copy of the script on those > > > machines -- or we can change 'make regtest' run all of them. I'm leaning > > > towards the latter, as an incentive to get the exp-tools' tests working > > > as well as possible. Julian, what do you think? > > I'd say, for now change the script to run make exp-regtest too. > Tom's machines will auto-update and I'll change mine to do whatever > Tom's are doing. After 3.3.0 is done we can merge make exp-regtest > into standard make regtest. Now that 3.3.0 has been released, can the aforementioned change please be implemented ? I recommend to run the exp-drd regression tests only on x86 and x86_64, and not on ppc or AIX. Bart. |
|
From: Nicholas N. <nj...@cs...> - 2008-02-10 22:21:53
|
On Mon, 7 Jan 2008, Bart Van Assche wrote: >> After 3.3.0 is done we can merge make exp-regtest into standard make >> regtest. > > Now that 3.3.0 has been released, can the aforementioned change please > be implemented ? I recommend to run the exp-drd regression tests only > on x86 and x86_64, and not on ppc or AIX. Done. Sorry it took so long, I lost track of this email thread. Nick |
|
From: Bart V. A. <bar...@gm...> - 2008-02-11 19:16:21
|
On Feb 10, 2008 11:21 PM, Nicholas Nethercote <nj...@cs...> wrote: > On Mon, 7 Jan 2008, Bart Van Assche wrote: > > >> After 3.3.0 is done we can merge make exp-regtest into standard make > >> regtest. > > > > Now that 3.3.0 has been released, can the aforementioned change please > > be implemented ? I recommend to run the exp-drd regression tests only > > on x86 and x86_64, and not on ppc or AIX. > > Done. Sorry it took so long, I lost track of this email thread. No problem :-) Bart. |