Is it failing on Windows only? It could be an issue with the
granularity of timestamps. I think we use more accurate timestamps on
Linux nowadays, whereas Windows is still using 1 second granularity. If
that's the case, it probably needs a "sleep 1" somewhere. If the "sleep
1" didn't help, perhaps it wasn't in the right place - it needs to be
before the second set of echo commands.
On the subject of naming the test, we want two pieces of information in
the name:
- what it tests
- the ticket number, if it's a regression test
So rather than just renaming it to the ticket number (which would lose
the information that it's a recomp test), I suggest moving all the
recomp tests into their own directory first. It's really useful to have
all the tests for a component together, so you can run them quickly when
you're working on that component.
Cheers,
Simon
On 19/09/2013 08:57, Gabor Greif wrote:
On 9/19/13, Simon Peyton-Jones <simo...@microsoft.com> wrote:
I lied. Whether or not the 'sleep' is there, the test usually fails, but
occasionally (one time in ten or so) succeeds.
I am the originator of the bug report, and could consistently
reproduce it in a bigger project. Unfortunately I have no time
currently to dive into the issue, but I have it earmarked for later.
Cheers,
Gabor
SImon
| -----Original Message-----
| From: ghc-devs [mailto:ghc-devs-boun...@haskell.org] On Behalf Of Simon
| Peyton-Jones
| Sent: 19 September 2013 12:50
| To: Joachim Breitner; ghc-devs@haskell.org
| Subject: RE: recomp013
|
| that makes it work.!
|
| | -----Original Message-----
| | From: Joachim Breitner [mailto:m...@joachim-breitner.de]
| | Sent: 19 September 2013 12:41
| | To: Simon Peyton-Jones; ghc-devs@haskell.org
| | Subject: Re: recomp013
| |
| | Dear Simon,
| |
| | Am Donnerstag, den 19.09.2013, 11:33 +0000 schrieb Simon Peyton-Jones:
| | > =====> recomp013(normal) 89 of 114 [0, 0, 0]
| | > cd ./recomp013 && $MAKE -s --no-print-directory recomp013
| | </dev/null >recomp013.run.stdout 2>recomp013.run.stderr
| | > Actual stdout output differs from expected:
| | > --- ./recomp013/recomp013.stdout 2013-09-18 12:38:22.000000000
| +0100
| | > +++ ./recomp013/recomp013.run.stdout 2013-09-19 12:15:22.611120935
| | +0100
| | > @@ -3,5 +3,3 @@
| | > [2 of 3] Compiling B ( B.hs, B.o )
| | > [3 of 3] Compiling C ( C.hs, C.o )
| | > second run
| | > -[2 of 3] Compiling B ( B.hs, B.o )
| | > -[3 of 3] Compiling C ( C.hs, C.o ) [B changed]
| | > *** unexpected failure for recomp013(normal)
| |
| | It works here. So that means either two things:
| | * that the bug described in #8247 is actually present in HEAD, but
| | only under certain circumstances, or
| | * that there is a timing issue with the update of B.hs.
| |
| | The latter seems to be more likely. I notice that other recomp-tests
| | have a "sleep 1" before updating the file: If you insert that in
| | recomp013/Makefile, does it go through for you?
| |
| | Greetings,
| | Joachim
| |
| | --
| | Joachim “nomeata” Breitner
| | m...@joachim-breitner.de • http://www.joachim-breitner.de/
| | Jabber: nome...@joachim-breitner.de • GPG-Key: 0x4743206C
| | Debian Developer: nome...@debian.org
| _______________________________________________
| ghc-devs mailing list
| ghc-devs@haskell.org
| http://www.haskell.org/mailman/listinfo/ghc-devs
_______________________________________________
ghc-devs mailing list
ghc-devs@haskell.org
http://www.haskell.org/mailman/listinfo/ghc-devs
_______________________________________________
ghc-devs mailing list
ghc-devs@haskell.org
http://www.haskell.org/mailman/listinfo/ghc-devs
_______________________________________________
ghc-devs mailing list
ghc-devs@haskell.org
http://www.haskell.org/mailman/listinfo/ghc-devs
