alpmestan (Alp Mestanogullari)Administrator
User

Projects

User does not belong to any projects.

User Details

User Since
Jun 5 2014, 10:16 AM (232 w, 2 d)
Roles
Administrator

Recent Activity

Yesterday

alpmestan added a comment to D5270: Hadrian: merge sections in profiling _p.a to .p_o for ghci.

There's a diff that bumps the Cabal submodule off the 2.4 branch, to track master: https://phabricator.haskell.org/D5329

Sat, Nov 17, 3:55 AM

Fri, Nov 16

alpmestan created D5345: fix T15898.
Fri, Nov 16, 2:25 PM

Thu, Nov 15

alpmestan created D5341: bump haddock submodule.
Thu, Nov 15, 4:24 PM

Wed, Nov 14

alpmestan closed D4989: Adjust normal runtimes for nofib along with related changes..
Wed, Nov 14, 3:46 PM
alpmestan committed rNOFIB08cc9b6b2c7f: Adjust normal runtimes for nofib along with related changes (authored by AndreasK).
Adjust normal runtimes for nofib along with related changes
Wed, Nov 14, 3:46 PM
alpmestan added a comment to D4989: Adjust normal runtimes for nofib along with related changes..

Merged in 08cc9b6b2c7f7fdaaaf80361ab84a501f0a573c5. That somehow didn't close the diff automatically, and I can't mark the diff as closed myself.

Wed, Nov 14, 11:21 AM
alpmestan accepted D5335: Building GHC with hadrian on FreeBSD.
Wed, Nov 14, 9:43 AM
alpmestan committed rGHC89fa34ecd326: hadrian: build ghc-iserv-dyn (authored by DavidEichmann).
hadrian: build ghc-iserv-dyn
Wed, Nov 14, 8:49 AM
alpmestan closed D5327: hadrian: build ghc-iserv-dyn in addition to ghc-iserv and ghc-iserv-prof, as it is required for 10+ tests.
Wed, Nov 14, 8:49 AM

Mon, Nov 12

alpmestan committed rGHC86ee74dc999d: compareByPreference: handle the integer-gmp vs -simple case (authored by alpmestan).
compareByPreference: handle the integer-gmp vs -simple case
Mon, Nov 12, 10:58 AM
alpmestan closed D5266: compareByPreference: handle the case where the two packages have a different name.
Mon, Nov 12, 10:58 AM
alpmestan accepted D5327: hadrian: build ghc-iserv-dyn in addition to ghc-iserv and ghc-iserv-prof, as it is required for 10+ tests.

Modulo the typo in the comment, looks good to me.

Mon, Nov 12, 5:02 AM

Mon, Nov 5

alpmestan added a comment to D5266: compareByPreference: handle the case where the two packages have a different name.

Done.

Mon, Nov 5, 1:41 PM
alpmestan updated the diff for D5266: compareByPreference: handle the case where the two packages have a different name.
  • add reference to integer library note in PrelNames
Mon, Nov 5, 1:41 PM

Thu, Nov 1

alpmestan updated the diff for D5289: Add a RTS option -xp to load PIC object anywhere in address space.

(try pushing this patch to the staging repo)

Thu, Nov 1, 11:40 AM

Wed, Oct 31

alpmestan added a comment to D5281: Dynamically linked ghc uses relative path to shared objects..

LGTM. It's a little annoying that you have to define makeRelativeNoSysLink and handle all those things in Hadrian itself. Perhaps could we at least throw that function's definition in a dedicated module (or in Utilities or something). And indeed what about Windows?

Wed, Oct 31, 2:30 AM

Mon, Oct 29

alpmestan abandoned D5230: rts/Linker.c: define the 'mmap_again' label only for x86_64.

https://phabricator.haskell.org/D5273 is taking over.

Mon, Oct 29, 11:47 AM
alpmestan accepted D5273: Fix unused label on non x86_64 platforms.

Hmm, I might have jumped to some conclusions too quickly, my bad. I'm seeing 300+ test failures here (i386 Circle CI build of a recent master commit of GHC), which is about the amount I saw as well. So I guess either of our diffs should be merged as they don't cause any (additional) problem down the road. Let's go with yours.

Mon, Oct 29, 11:46 AM
alpmestan added a comment to D5273: Fix unused label on non x86_64 platforms.

A whole lot of test failures, see https://circleci.com/gh/ghc/ghc-diffs/484;

Mon, Oct 29, 5:39 AM
alpmestan added a comment to D5266: compareByPreference: handle the case where the two packages have a different name.

@bgamari How about this?

Mon, Oct 29, 5:32 AM
alpmestan updated the diff for D5266: compareByPreference: handle the case where the two packages have a different name.
  • only handle different packages when they're integer-{gmp, simple}
Mon, Oct 29, 5:31 AM
alpmestan added a comment to D5266: compareByPreference: handle the case where the two packages have a different name.

I answered a bit too quickly, yesterday. integer-wired-in doesn't get passed to compareByPreference. I assumed you meant "integer-wired-in replacements", so to speak. The reason why we end up comparing two different packages is that just a little earlier in the pipeline, we open that possibility by saying that there are two packages that can match integerUnitId. So I'll proceed accordingly, by introducing a clause not for the general case of two different packages, but specifically for the case where one is integer-gmp and the other is integer-simple.

Mon, Oct 29, 4:33 AM

Sun, Oct 28

alpmestan planned changes to D5266: compareByPreference: handle the case where the two packages have a different name.

@bgamari This sounds perfectly reasonable and will address our problem. I'll tweak my patch accordingly ASAP.

Sun, Oct 28, 1:30 PM
alpmestan added a comment to D5273: Fix unused label on non x86_64 platforms.

Hmm, I tried the same thing in https://phabricator.haskell.org/D5230 but it causing problems.

Sun, Oct 28, 10:20 AM

Sat, Oct 27

alpmestan added inline comments to D5270: Hadrian: merge sections in profiling _p.a to .p_o for ghci.
Sat, Oct 27, 2:50 PM

Fri, Oct 26

alpmestan added inline comments to D5266: compareByPreference: handle the case where the two packages have a different name.
Fri, Oct 26, 7:25 AM
alpmestan created D5266: compareByPreference: handle the case where the two packages have a different name.
Fri, Oct 26, 5:53 AM

Thu, Oct 25

alpmestan added a comment to D5255: hadrian: build ghc-iserv-prof in addition to ghc-iserv, as it is required for 10+ tests.

@alpmestan Great! Is there a way to migrate our AppVeyor and Travis scripts too?

Could you also explain me how it works in general: will a Hadrian build be triggered for every Phab differential? Where can reviewers find corresponding build results?

Thu, Oct 25, 7:01 AM
alpmestan added a comment to D5255: hadrian: build ghc-iserv-prof in addition to ghc-iserv, as it is required for 10+ tests.

The build passed.

Thu, Oct 25, 4:22 AM
alpmestan added a comment to D5255: hadrian: build ghc-iserv-prof in addition to ghc-iserv, as it is required for 10+ tests.

@snowleopard I just launched a hadrian build on Circle CI: https://circleci.com/gh/ghc/ghc-diffs/513

Thu, Oct 25, 3:44 AM

Wed, Oct 24

alpmestan added inline comments to D5255: hadrian: build ghc-iserv-prof in addition to ghc-iserv, as it is required for 10+ tests.
Wed, Oct 24, 1:10 PM
alpmestan added a comment to D5255: hadrian: build ghc-iserv-prof in addition to ghc-iserv, as it is required for 10+ tests.

@alpmestan Thank you, this looks good to me. However, have we lost all CI support with the move from GitHub? Can we fix the CI first?

Wed, Oct 24, 12:16 PM
alpmestan created D5255: hadrian: build ghc-iserv-prof in addition to ghc-iserv, as it is required for 10+ tests.
Wed, Oct 24, 7:34 AM

Tue, Oct 23

GitHub <noreply@github.com> committed rGHC051564be1f98: update GHC versions used on travis (authored by alpmestan).
update GHC versions used on travis
Tue, Oct 23, 4:41 PM
GitHub <noreply@github.com> committed rGHCdb3dbb614525: Merge pull request #700 from snowleopard/alp/ghc-versions-ci (authored by alpmestan).
Merge pull request #700 from snowleopard/alp/ghc-versions-ci
Tue, Oct 23, 4:41 PM

Oct 19 2018

alpmestan accepted D5243: testsuite: Mark T15349 as broken in the ghci way.
Oct 19 2018, 3:54 AM
alpmestan accepted D5242: testsuite: EtaExpandLevPoly now passes in profiled ways.
Oct 19 2018, 3:21 AM

Oct 18 2018

alpmestan added a comment to D5230: rts/Linker.c: define the 'mmap_again' label only for x86_64.

https://circleci.com/gh/ghc/ghc-diffs/484 is the build for my latest update. So, let me sum it up: on i686, not having this mmap_again label there is enough to make 300+ tests fail, even though (still on i686) we _never_ jump to it. I'll have to try and reproduce this on a small isolated example to figure out what's going on at the assembly level, none of this makes sense to me.

Oct 18 2018, 3:04 AM

Oct 17 2018

alpmestan updated the diff for D5230: rts/Linker.c: define the 'mmap_again' label only for x86_64.
  • only define mmap_again label when on x86_64
Oct 17 2018, 4:13 PM
alpmestan added a comment to D5230: rts/Linker.c: define the 'mmap_again' label only for x86_64.

I'm a bit lost. Is the problem that mmap_again is considered to be unused since the compiler has shown all gotos are unreachable?

Oct 17 2018, 9:54 AM
alpmestan added a comment to D5230: rts/Linker.c: define the 'mmap_again' label only for x86_64.

There's an i686 (Circle CI) build running for this diff here: https://circleci.com/gh/ghc/ghc-diffs/479

Oct 17 2018, 8:19 AM
alpmestan updated the diff for D5230: rts/Linker.c: define the 'mmap_again' label only for x86_64.
  • previous fix was wrong, let's try an explicit goto to silence the error
Oct 17 2018, 8:08 AM
alpmestan added a comment to D5230: rts/Linker.c: define the 'mmap_again' label only for x86_64.

The i686 build now fails, complaining that most of the variables we define are unused.

Oct 17 2018, 7:59 AM
alpmestan added a comment to rGHC104599f3f157: Add a RTS option -xp to load PIC object anywhere in address space.

For the record, this broke the i386 build: https://phabricator.haskell.org/harbormaster/build/54657/

Oct 17 2018, 4:35 AM
alpmestan updated the summary of D5230: rts/Linker.c: define the 'mmap_again' label only for x86_64.
Oct 17 2018, 4:31 AM
Herald added a reviewer for D5230: rts/Linker.c: define the 'mmap_again' label only for x86_64: erikd.
Oct 17 2018, 4:31 AM

Oct 10 2018

alpmestan added a comment to D5216: testsuite: Add tests for #15270.

@bgamari The tests that you added seem to be failing on OS X, judging from https://phabricator.haskell.org/harbormaster/log/view/305029/

Oct 10 2018, 4:28 AM

Oct 9 2018

alpmestan added a comment to D5215: update amd64 linux/osx stats for haddock.base perf test.

Landed per Ben's request since this failure was making a lot of (most?) CI runs fail.

Oct 9 2018, 5:57 AM
alpmestan committed rGHC4eeeb51d5f51: update amd64 linux/osx stats for haddock.base perf test (authored by alpmestan).
update amd64 linux/osx stats for haddock.base perf test
Oct 9 2018, 5:43 AM
alpmestan closed D5215: update amd64 linux/osx stats for haddock.base perf test.
Oct 9 2018, 5:43 AM

Oct 8 2018

alpmestan accepted D5216: testsuite: Add tests for #15270.
Oct 8 2018, 5:15 PM
alpmestan updated the diff for D5215: update amd64 linux/osx stats for haddock.base perf test.
  • tab -> spaces
Oct 8 2018, 3:26 PM
alpmestan created D5215: update amd64 linux/osx stats for haddock.base perf test.
Oct 8 2018, 3:22 PM

Oct 2 2018

alpmestan accepted D5191: circleci: Run slowtest with multiple threads.

Oh, we were not generating the XML report that the Circle CI user interface picks up to give us the failing tests either? Interesting.

Oct 2 2018, 1:43 AM

Oct 1 2018

alpmestan added a comment to D5166: rts.cabal.in: add more flags to control the extra flavours (for hadrian).

Yes, and that's still where this is handled for the Make build system, since it doesn't use the cabal file's information.

Oct 1 2018, 2:22 PM

Sep 28 2018

alpmestan added a comment to rGHC900c47f88784: rts/Printer.c: always define the findPtr symbol.

Hmm, this solution does seem inferior to the one you suggest, I was not aware of:

Sep 28 2018, 4:43 AM

Sep 20 2018

alpmestan added a comment to D4880: Fix for built-in Natural literals.

hadrian doesn't need this patch anymore indeed, and I believe this was the only motivation for this patch -- can you confirm @hsyl20?

Sep 20 2018, 6:55 PM
alpmestan added a comment to D5168: Always check the relocation value for x86_64.

Great! I believe this would have helped me during my investigation from Trac #15570.

Sep 20 2018, 10:42 AM
alpmestan created D5166: rts.cabal.in: add more flags to control the extra flavours (for hadrian).
Sep 20 2018, 8:53 AM

Sep 12 2018

alpmestan added a comment to D5146: ghc-bin.cabal.in: add a 'threaded' flag for hadrian.

I can confirm that T8242 passes now, with the accompanying hadrian PR (here):

Sep 12 2018, 6:13 PM
alpmestan added a comment to D5146: ghc-bin.cabal.in: add a 'threaded' flag for hadrian.

I have a build running, and it'll run the test at the end. I'll report back once it's all over.

Sep 12 2018, 5:00 PM
alpmestan created D5146: ghc-bin.cabal.in: add a 'threaded' flag for hadrian.
Sep 12 2018, 4:56 PM

Sep 11 2018

alpmestan added a comment to D5138: rts/Printer.c: always define the findPtr symbol.

Looks like that didn't break anything for the make build system either, judging from the harbormaster builds.

Sep 11 2018, 3:45 AM

Sep 10 2018

alpmestan created D5142: rts.cabal.in: advertise new default profiling ways (introduced in D5140) for hadrian.
Sep 10 2018, 10:42 AM
alpmestan added a comment to D5138: rts/Printer.c: always define the findPtr symbol.

@simonmar I implemented both of your suggestions. Thanks!

Sep 10 2018, 5:33 AM
alpmestan updated the diff for D5138: rts/Printer.c: always define the findPtr symbol.
  • remove findPtr hack in rts/package.conf.in, give an helpful error message when not running a debugged RTS
  • silence unused argument warning/error
Sep 10 2018, 5:31 AM

Sep 7 2018

alpmestan updated the diff for D5138: rts/Printer.c: always define the findPtr symbol.
  • address unused parameter error
Sep 7 2018, 12:35 PM
alpmestan added a comment to D5138: rts/Printer.c: always define the findPtr symbol.

@simonmar Absolutely (see e.g the discussion here). In fact, you ended up "recommending" this approach yourself in D4846 , so I just went ahead and implemented it. :)

Sep 7 2018, 11:43 AM
alpmestan created D5138: rts/Printer.c: always define the findPtr symbol.
Sep 7 2018, 7:50 AM

Sep 4 2018

alpmestan created D5127: testsuite/mk/boilerplate.mk: make CHECK_API_ANNOTATIONS and CHECK_PPR overridable.
Sep 4 2018, 5:50 AM

Aug 31 2018

alpmestan accepted D5119: Skip eventlog tests in GHCi way.
Aug 31 2018, 4:57 AM

Aug 27 2018

alpmestan updated the Trac tickets for D5103: tests: increase (compile) timeout multiplier for T13701 and MultiLayerModules.
Aug 27 2018, 3:35 PM
alpmestan added a comment to D5103: tests: increase (compile) timeout multiplier for T13701 and MultiLayerModules.

Only T3171 left to fix (https://circleci.com/gh/ghc/ghc-diffs/270)

Aug 27 2018, 3:35 PM
alpmestan updated the diff for D5103: tests: increase (compile) timeout multiplier for T13701 and MultiLayerModules.
  • bigger multiplier for MultiLayerModules
Aug 27 2018, 1:13 PM
alpmestan added a comment to D5106: make iToBase62's inner loop stricter in one of its arguments.

And here's the ticket with the self-contaiend reproducer: https://ghc.haskell.org/trac/ghc/ticket/15570#ticket

Aug 27 2018, 1:13 PM
alpmestan added a comment to D5103: tests: increase (compile) timeout multiplier for T13701 and MultiLayerModules.

I made some more tweaks and triggered a new Circle CI build for i386: https://phabricator.haskell.org/harbormaster/build/52342/

Aug 27 2018, 6:29 AM
alpmestan updated the diff for D5103: tests: increase (compile) timeout multiplier for T13701 and MultiLayerModules.
  • tweak timeout multipliers for bkpcabal01 and T13701(again)
Aug 27 2018, 6:27 AM

Aug 24 2018

alpmestan created D5106: make iToBase62's inner loop stricter in one of its arguments.
Aug 24 2018, 4:35 PM
alpmestan planned changes to D5103: tests: increase (compile) timeout multiplier for T13701 and MultiLayerModules.

https://circleci.com/gh/ghc/ghc-diffs/256#build-parameters/containers/0: MultiLayerModules passed but another one is failing, and T13701 is still not passing.

Aug 24 2018, 10:03 AM
alpmestan added a comment to D5103: tests: increase (compile) timeout multiplier for T13701 and MultiLayerModules.

We'll eventually get the CircleCI build log url and more.

Aug 24 2018, 7:04 AM
alpmestan created D5103: tests: increase (compile) timeout multiplier for T13701 and MultiLayerModules.
Aug 24 2018, 7:04 AM
alpmestan created D5102: fix -ddump-asm description.
Aug 24 2018, 5:50 AM

Aug 22 2018

alpmestan added a comment to D4990: Invert FP conditions to eliminate the explicit NaN check..

That was my suspicion, thanks for clarifying!

Aug 22 2018, 12:00 PM
alpmestan accepted D5083: base: showEFloat: Handle negative precisions the same of zero precision.
Aug 22 2018, 11:20 AM
alpmestan accepted D4990: Invert FP conditions to eliminate the explicit NaN check..

Out of curiosity, if this is a _performance_ optimisation, do we have any numbers we can leave here for posterity?

Aug 22 2018, 10:44 AM
alpmestan requested changes to D4717: [WIP] CoreToStg: Try treating String unpackings as single-entry.

The patch looks good, but Note [String unpack closures are non-updateable] has yet to be written.

Aug 22 2018, 8:28 AM
alpmestan added a comment to D4739: Fix "redundant constraint" warnings when meets functional dependencies.

The only CI failures are on OS X and they're stats failures. I'm no typechecker expert but the feedback got addressed and the test included with this patch is indeed passing. Any objection to merging this fix?

Aug 22 2018, 5:50 AM

Aug 4 2018

alpmestan accepted D5043: circleci: Force BUILD_SPHINX_*=YES.
Aug 4 2018, 12:27 AM

Aug 2 2018

alpmestan accepted D5037: circleci: Don't build validate-x86_64-linux-debug unregisterised.

Ha! That explains why I was seeing so many common failures between the unreg and debug validation jobs in Trac #15466 and Trac #15467. Thanks!

Aug 2 2018, 10:23 AM
alpmestan added a comment to D5034: Fix precision of asinh/acosh/atanh by making them primops.

Can you checkout 2a3175f886b86697194256f55c9487b7cfb4dc92 in libraries/Cabal/, make sure you don't have dirty files in that sub-directory, and then from the top of the GHC source tree git add libraries/Cabal && git commit ... && arc diff ?

Aug 2 2018, 1:30 AM
alpmestan added a comment to D5034: Fix precision of asinh/acosh/atanh by making them primops.

It should be enough to checkout whatever commits master points to for those two submodules, git add libraries/Cabal libraries/terminfo and then git commit -m ... and arc diff.

Aug 2 2018, 1:16 AM

Aug 1 2018

alpmestan created D5033: use *test instead of *slowtest for llvm validation on Circle CI.
Aug 1 2018, 1:03 PM
alpmestan abandoned D5024: bump Cabal submodule for hadrian issue #634.

Done in another commit.

Aug 1 2018, 11:26 AM
alpmestan created D5031: fix timeout related i686 validation issues.
Aug 1 2018, 10:53 AM

Jul 31 2018

alpmestan accepted D5027: users-guide: Document default +RTS -I value.
Jul 31 2018, 1:43 AM
alpmestan accepted D5026: base: Fix documentation of System.Environment.Blank.
Jul 31 2018, 1:33 AM

Jul 30 2018

alpmestan created D5024: bump Cabal submodule for hadrian issue #634.
Jul 30 2018, 10:50 AM
alpmestan accepted D5015: Use -fobject-code in the GHCi script for loading GHC.

LGTM. Yeah -fobject-code seems to be a better default. Just make sur you document the most efficient/handy (but more complicated) workflow somewhere visible like one of the "building ghc" pages on trac eventually, so that people can pick it up.

Jul 30 2018, 7:04 AM

Jul 27 2018

alpmestan updated the diff for D4229: ghc, ghc-pkg: use getExecutablePath on Windows when base >= 4.11.0.
  • Merge branch 'master' into wip/alp/14483-getexecutablepath-windows
Jul 27 2018, 4:34 AM

Jul 26 2018

alpmestan added a comment to D4229: ghc, ghc-pkg: use getExecutablePath on Windows when base >= 4.11.0.

I didn't get around to figuring out the source of the Windows CI troubles. I can put this patch back on my radar and look into this in the upcoming days.

Jul 26 2018, 6:04 PM