Jun 11 2019
Dec 10 2018
Throw a warning if -Wunused-packages is used without -fforce-recomp.
- Mention -fforce-recomp in the user guide
Dec 3 2018
Oops, I missed notifications for this Diff somehow, sorry.
@bgamari so what actions should I perform to get it merged? Suggest -fforce-recomp in user guide? Anything else?
Nov 8 2018
@mpickering On other hand, why should we force user to recompile all the code if she can get results immediately, even if less occurate? So I'm -0.5 for enabling -fforce-recomp by default, I think it is enough to suggest it in the user guide. @bgamari, what do you think about that?
Nov 7 2018
@mpickering I understand your concern, but IMO it is completely clear what the patch does in case of a clean build. It is not clear why GHC loads all packages when doing partial recompilation, and I need someone's help to figure it out. Do you have any idea? My best guess is that packages are necessary for linking executable/library, but I'm not familiar with this part of GHC. As you probably know, I contribute only occasionally, and my knowledges are sparse.
Nov 6 2018
@bgamari I added a release note.
Nov 1 2018
Ok, I fixed the issue with wired-in packages. I inlined the matching function from Packages and modified it. Not sure whether the original function should be fixed instead.
- Fix wired-in package handling
I wasn't able to quickly find the reason for the false positivities, so lets change status to prevent accidental merge.
@simonpj I added a test and reworded the message:
The following packages were specified via -package or -package-id flags, but were not needed for compilation:
- Change message, add test
Oct 30 2018
@mpickering To my surprice, it seems to work when run the second time. I'm not sure why. Anyway, the idea is to use this flag on a clean code base.
This probably requires tests yet.
May 22 2017
LGTM. @hvr, thank you for taking care of of this patch!
May 19 2017
- Handle also file targets, converted from module targets
Jan 18 2017
Jan 16 2017
@simonpj My English writing skills are relatively poor, sorry. Please let me know if the Note is not clear enough.
Jan 15 2017
T12234 failure looks unrelated for me, I see it in other builds. But recomp001 failure on osx is puzzling, does anyone has an idea what is going here? I can't reproduce it locally (linux).
T12234 failure looks unrelated for me, I see it in other builds. Everything else is OK.
Feb 27 2016
Thank you, such left-overs are confusing when reading code the first time.
Feb 15 2016
Dec 8 2015
The following program crashes for me:
Aug 26 2015
I didn't find a way to make a test for hp2ps. Probably it doesn't worth it.
Don't use a global variable
I just found that the test case is not suitable for the issue because package key for 'random' package (used in the test case) may change in future, so it will not start with a digit any more. We need a special test case for the issue.
- Your commit message is difficult to parse.
Aug 25 2015
Feb 15 2015
Feb 12 2015
Dec 8 2014
Dec 7 2014
update test output
add both the original and specialized types into error context
Dec 6 2014
@austin what about the next:
@austin are you sure but is necessary here? It suggests that there is a mismatch between number of arguments and function arity, but that is not true. This two lines provide an additional context in case the inferred is specialized in unexpected way, e.g. when user missed do or parenthesis.
Dec 4 2014
Nov 15 2014
Oct 24 2014
Oct 22 2014
Oct 20 2014
Oct 19 2014
Regarding the manual, didn't you extend the foreign import syntax to allow users to specify a struct return type?
Oct 18 2014
Address (most of) issues @tibbe raised
@tibbe Thanks you for you comments!
Oct 16 2014
Oct 13 2014
@nomeata do you mean rGHCf3ae936e04dda1b28b0fa8f4c65a99407164ee99?
Hmm... it doesn't seem to touch T4801, at least in master the values are still different: https://phabricator.haskell.org/diffusion/GHC/browse/master/testsuite/tests/perf/compiler/all.T;f3ae936e04dda1b28b0fa8f4c65a99407164ee99$184
Oct 8 2014
32bit mac os support
Oct 7 2014
Sorry I incidentally pressed submit before the comment finished.
Not the code doesn't use red zone (which) is linux specific.
Don't use red zone when
Oct 6 2014
I tested it on mac os X x86_64 -- just works. Mac os on x86 is probably dead, I don't have access to it.
Oct 3 2014
Implement windows x86 support
Sep 27 2014
oops, I messed up with arc...
Add Notes re calling convention, few haddock comments, reorder functions, typo