I'm glad you like it! When printing *lists*, there's also pprDeeperList (or
something like it). Often you want to cut off after a certain number of
elements.
Simon
| -Original Message-
| From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of pepe
| Sent: 15 November 2007 09:56
Build description = STABLE on i386-unknown-linux
(cam-02-unx.europe.corp.microsoft.com)
Build location= /playpen/ghc/nightly/STABLE-cam-02-unx
Build config file = /home/simonmar/nightly/site/msrc/conf-STABLE-cam-02-unx
Nightly build started on cam-02-unx at Thu Nov 15 19:00:01 GMT 2007.
On Thu, Nov 15, 2007 at 09:00:18AM +, Simon Peyton-Jones wrote:
> Simon, Ian
>
> We don't have a good way to test for performance regressions, and we really,
> really ought to.
>
> Printing heap allocated, and regarding the test as failed if it get more than
> 2% worse might be a simple way
Wed Nov 14 08:45:25 PST 2007 Simon Marlow <[EMAIL PROTECTED]>
* FIX #1840: outofmem and outofmem2 need ulimit settings to avoid filling up
swap
M ./tests/ghc-regress/rts/Makefile +14
M ./tests/ghc-regress/rts/all.T -5 +4
A ./tests/ghc-regress/rts/outofmem.stdout
A ./tests/ghc-r
On Thu, 2007-11-15 at 02:14 +, Claus Reinke wrote:
> substring matching can optionally ignore cases to avoid the
> rather unpredictable capitalisation of packages.
If this is up for discussion, it'd be useful if ghc-pkg did not allow
the same version of packages that differ only in
Build description = HEAD on i386-unknown-linux
(cam-02-unx.europe.corp.microsoft.com)
Build location= /playpen/ghc/nightly/HEAD-cam-02-unx
Build config file = /home/simonmar/nightly/site/msrc/conf-HEAD-cam-02-unx
Nightly build started on cam-02-unx at Thu Nov 15 19:30:01 GMT 2007.
checki
unfortunately, disabling mingw globbing is not the solution,
as it leaves no globbing at all under cmd.exe:
C:\fptools\ghc\utils\ghc-pkg>tst-glob *
[".depend",".depend.bak",".ghc-pkg-inplace.hs.swp",".Main.hs.swp","ghc-pkg-inpla
ce.exe","ghc-pkg-inplace.hi","ghc-pkg-inplace.hs","ghc-pk
but that does seem to be a bug in ghc-pkg-inplace.hs,
so perhaps that should be fixed first, and then we could use '*' instead of '.' (even though it
isn't full globbing, '*' is only interpreted at string ends).
the problem is not specific to ghc-pkg-inplace, but is
shared by all ghc mingw prog
Thu Nov 15 08:23:57 PST 2007 Simon Marlow <[EMAIL PROTECTED]>
* outofmem2 passes on the stable branch
M ./tests/ghc-regress/rts/all.T -2 +1
___
Cvs-ghc mailing list
Cvs-ghc@haskell.org
http://www.haskell.org/mailman/listinfo/cvs-ghc
Wed Nov 14 08:39:45 PST 2007 Simon Marlow <[EMAIL PROTECTED]>
* FIX #1791 on stable branch
The 6.8 branch needed this fix in another place
M ./rts/sm/Storage.c +8
___
Cvs-ghc mailing list
Cvs-ghc@haskell.org
http://www.haskell.org/mailman/listi
Wouldn't * be more conventional than .? * means "any substring" in
filename globbing, so most people will be familiar with that usage. I
suppose it would have to be quoted, but I'd pay that penalty. Anyone else
have any thoughts?
that was my first tendency as well, but unfortunately, the
qu
ThreadDelay is a perfectly reasonable test: it tests that threadDelay
doesn't wait for *less* time than was requested. Clearly something is
wrong on your machine with either threadDelay or the way time is measured,
but we haven't seen the same symptoms elsewhere so it's hard to diagnose.
Still
On Thu, 2007-11-15 at 13:26 +, Simon Marlow wrote:
> Actually there could be a user package DB, and that could cause this to go
> wrong (but only if the user has some of the standard packages registered in
> their DB, which is fairly unlikely I suppose).
Core packages are upgradable :-) Tha
On 14/11/2007, at 20:43, Claus Reinke wrote:
Claus, is this expected to work with 6.8 ?
There is no ':show languages' option there, so the test fails.
i should have said STABLE branch, i guess.
The test passes in STABLE branch. I'll be pushing asap.
but this raises a question for those
Thu Nov 15 05:15:17 PST 2007 Simon Marlow <[EMAIL PROTECTED]>
* add test for #1679
A ./tests/ghc-regress/ccall/should_run/1679.hs
A ./tests/ghc-regress/ccall/should_run/1679.stdout
M ./tests/ghc-regress/ccall/should_run/all.T +3
___
Cvs-g
Thu Nov 15 05:16:35 PST 2007 Simon Marlow <[EMAIL PROTECTED]>
* FIX #1679: crash on returning from a foreign call
We forgot to save a pointer to the BCO over the foreign call. Doing
enough allocation and GC during the call could provoke a crash.
M ./rts/Interpreter.c -1 +16
__
Duncan Coutts wrote:
On Thu, 2007-11-15 at 03:11 -0800, Simon Marlow wrote:
Thu Nov 15 02:32:49 PST 2007 Simon Marlow <[EMAIL PROTECTED]>
* Avoid the use of unversioned package dependencies
Fortunately "ghc-pkg list $pkg --simple-output" is a good way to add
the version number.
You're l
On Thu, 2007-11-15 at 03:11 -0800, Simon Marlow wrote:
> Thu Nov 15 02:32:49 PST 2007 Simon Marlow <[EMAIL PROTECTED]>
> * Avoid the use of unversioned package dependencies
> Fortunately "ghc-pkg list $pkg --simple-output" is a good way to add
> the version number.
You're lucky that you've
On Thu, 2007-11-15 at 10:24 +, Simon Marlow wrote:
> Claus Reinke wrote:
> > * fix #1839, #1463, by supporting ghc-pkg bulk queries with substring
> > matching
> >
> > - #1839 asks for a ghc-pkg dump feature, #1463 for the ability
> > to query the same fields in several packages at onc
Thu Nov 15 02:32:49 PST 2007 Simon Marlow <[EMAIL PROTECTED]>
* Avoid the use of unversioned package dependencies
Fortunately "ghc-pkg list $pkg --simple-output" is a good way to add
the version number.
M ./compiler/Makefile -1 +3
___
Cvs-ghc
Wed Nov 14 08:53:23 PST 2007 Simon Marlow <[EMAIL PROTECTED]>
* FIX #1596 (remove deprecated --define-name)
Also remove the old command-line syntax for ghc-pkg, which was not
documented. Do not merge.
M ./utils/ghc-pkg/Main.hs -139 +14
___
C
Wed Nov 14 08:10:44 PST 2007 Simon Marlow <[EMAIL PROTECTED]>
* FIX #1837: remove deprecated support for unversioned dependencies (do not
merge)
M ./utils/ghc-pkg/Main.hs -36 +1
___
Cvs-ghc mailing list
Cvs-ghc@haskell.org
http://www.haskell.
Claus Reinke wrote:
* fix #1839, #1463, by supporting ghc-pkg bulk queries with substring
matching
- #1839 asks for a ghc-pkg dump feature, #1463 for the ability
to query the same fields in several packages at once.
- this patch enables substring matching for packages in 'list',
'
Wed Nov 14 15:25:53 PST 2007 Pepe Iborra <[EMAIL PROTECTED]>
* Accept output
The test now actually passes
M ./tests/ghc-regress/ghci.debugger/scripts/print021.stdout -1 +1
___
Cvs-ghc mailing list
Cvs-ghc@haskell.org
http://www.haskell.or
Claus Reinke wrote:
btw, i've never seen a clean validate run on my system (win/xp), as
ThreadDelay001(normal) fails consistently.
Can you fix that or anybody else who uses that platform? (I am
making an effort to hunt down all Mac OS X breakages that I find.
For every platform, we need
Wed Nov 14 15:33:43 PST 2007 Pepe Iborra <[EMAIL PROTECTED]>
* Make pprNameLoc more robust in absence of loc information
M ./compiler/basicTypes/Name.lhs -1 +4
___
Cvs-ghc mailing list
Cvs-ghc@haskell.org
http://www.haskell.org/mailman/listinfo/c
Wed Nov 14 10:34:17 PST 2007 Pepe Iborra <[EMAIL PROTECTED]>
* Make the Term ppr depth aware
M ./compiler/ghci/RtClosureInspect.hs -29 +31
___
Cvs-ghc mailing list
Cvs-ghc@haskell.org
http://www.haskell.org/mailman/listinfo/cvs-ghc
Wed Nov 14 15:16:01 PST 2007 Pepe Iborra <[EMAIL PROTECTED]>
* Try to manage the size of the text rendered for ':show bindings'
M ./compiler/ghci/Debugger.hs -5 +18
M ./compiler/ghci/GhciMonad.hs -1 +8
M ./compiler/ghci/InteractiveUI.hs -19 +8
M ./compiler/ghci/RtClosureInspect.
Wed Nov 14 15:33:56 PST 2007 Pepe Iborra <[EMAIL PROTECTED]>
* wibble
M ./compiler/ghci/RtClosureInspect.hs -2 +2
___
Cvs-ghc mailing list
Cvs-ghc@haskell.org
http://www.haskell.org/mailman/listinfo/cvs-ghc
Wed Nov 14 10:12:33 PST 2007 Pepe Iborra <[EMAIL PROTECTED]>
* Use paragraph fill sep where possible
M ./compiler/ghci/RtClosureInspect.hs -4 +4
___
Cvs-ghc mailing list
Cvs-ghc@haskell.org
http://www.haskell.org/mailman/listinfo/cvs-ghc
The bunch of patches I just pushed make use of the ppr_deeper
primitives from Outputable, and now we cut the output at the depth(and
width) given by opt_PprUserLength.
I wasn't aware that this functionality was available in Outputable: it
is quite amazing what one can do with these in a few l
Thu Nov 15 00:42:42 PST 2007 [EMAIL PROTECTED]
* Make SpecConstr work again
In a typo I'd written env instead of env', and as a result RULES are
practically guaranteed not to work in a recursive group. This pretty
much kills SpecConstr in its tracks!
Well done Kenny Lu for spottin
Wed Nov 14 15:38:29 PST 2007 Pepe Iborra <[EMAIL PROTECTED]>
* Add test for #1892 - :info of a bkpt binding panics
M ./tests/ghc-regress/ghci.debugger/scripts/all.T +1
A ./tests/ghc-regress/ghci.debugger/scripts/break027.script
A ./tests/ghc-regress/ghci.debugger/scripts/break027.st
Wed Nov 14 15:24:45 PST 2007 Pepe Iborra <[EMAIL PROTECTED]>
* Adapt output to improvements in the TTerm pretty printer
M ./tests/ghc-regress/ghci.debugger/scripts/break024.stdout -15 +6
M ./tests/ghc-regress/ghci.debugger/scripts/print001.stdout -1 +2
M ./tests/ghc-regress/ghci.deb
Wed Nov 14 05:00:43 PST 2007 [EMAIL PROTECTED]
* simplify ghci024.py, fix -fprint-bind-result, -fno-ignore-breakpoints,
-fprint-bind-contents
- as ghc 6.9 flags are beginning to diverge, we need
a simpler way of specifying version-specific flags
or settings: have two entirely separ
Build results:
kahl G5 Gentoo Linux stable: pass
x86 Windows stable: pass
x86 Windows stable fast: pass pass pass fail (failed stage1) pass pass
x86-64 Linux stable: fail (failed darcs)
Old unexpected test passes:
1603 2 mnemosyne x86-64 Gentoo stable
New unexpected te
Build results:
x86-64 Linux head:pass
x86 Windows head: pass
x86 Windows head fast:pass pass pass pass pass pass
gabor head: pass
kahl G5 Gentoo Linux head:pass
mnemosyne x86-64 Gentoo head: pass
tnaur PPC OSX head: pass
tnaur x86 Linu
Simon, Ian
We don't have a good way to test for performance regressions, and we really,
really ought to.
Printing heap allocated, and regarding the test as failed if it get more than
2% worse might be a simple way. I'd like to have a primitive instruction count
too, but that seems harder. Me
38 matches
Mail list logo
