qnikst (Alexander Vershilov)
User

Projects

User does not belong to any projects.

User Details

User Since
Dec 3 2014, 4:33 AM (210 w, 6 d)

Recent Activity

Wed, Dec 12

qnikst updated the diff for D5369: RetainerProfiler: Update retainer profiler debugging.
  • actual rebase
Wed, Dec 12, 7:44 AM

Tue, Dec 11

qnikst updated the diff for D5369: RetainerProfiler: Update retainer profiler debugging.
  • rebase
Tue, Dec 11, 4:11 PM
qnikst added a comment to D5369: RetainerProfiler: Update retainer profiler debugging.

Can you rebase this, @qnikst?

Tue, Dec 11, 3:55 PM

Wed, Dec 5

qnikst updated the diff for D5351: Remove explicit recursion in retainer profiling (fixes #14758).
  • fixup
Wed, Dec 5, 10:36 AM
qnikst updated the diff for D5351: Remove explicit recursion in retainer profiling (fixes #14758).
  • Initialize posTypeObjects on the stack.
Wed, Dec 5, 10:21 AM
qnikst added a comment to D5351: Remove explicit recursion in retainer profiling (fixes #14758).

ok, the problem it seems that there is an actualPush function in Profiling.c, what is better to do, remove INLINE and rename the function or keep both?

Wed, Dec 5, 5:31 AM
qnikst added a comment to D5351: Remove explicit recursion in retainer profiling (fixes #14758).

@osa1 , @bgamari please take another look, the problem appears to be in the removing INLINE from the actual push. While I don't understand the reasons for that
I was able to find that while bisecting. In addition, I have renamed pushClosure to name the name unique in the project.

Wed, Dec 5, 5:27 AM
qnikst updated the diff for D5351: Remove explicit recursion in retainer profiling (fixes #14758).
  • Fixes.
Wed, Dec 5, 5:25 AM

Tue, Dec 4

qnikst added inline comments to D5351: Remove explicit recursion in retainer profiling (fixes #14758).
Tue, Dec 4, 11:38 AM
qnikst added inline comments to D5351: Remove explicit recursion in retainer profiling (fixes #14758).
Tue, Dec 4, 11:33 AM

Fri, Nov 30

qnikst updated the diff for D5351: Remove explicit recursion in retainer profiling (fixes #14758).
  • Remove INLINE on actualPush function
Fri, Nov 30, 2:03 PM
qnikst added a comment to D5351: Remove explicit recursion in retainer profiling (fixes #14758).

@bgamari please take another look, I've updated documentation a bit and have added a general information.
I was not able to dig why do we keep mutable objects as a retainer, but I have decided not to change semantics there.

Fri, Nov 30, 1:34 PM
qnikst updated the diff for D5351: Remove explicit recursion in retainer profiling (fixes #14758).
  • Add a note about retainers.
Fri, Nov 30, 1:32 PM

Thu, Nov 22

qnikst added a comment to D5351: Remove explicit recursion in retainer profiling (fixes #14758).

In general this file is terribly under-documented. In particular, there is no discussion of what the various stack entry types mean.

Thu, Nov 22, 2:14 PM
qnikst added a comment to D5351: Remove explicit recursion in retainer profiling (fixes #14758).

The implementation looks fine but I recall we discussed adding a note. What happened to this? It is badly needed.

Thu, Nov 22, 1:02 PM

Wed, Nov 21

qnikst created D5369: RetainerProfiler: Update retainer profiler debugging.
Wed, Nov 21, 2:45 PM

Tue, Nov 20

qnikst added a comment to D5351: Remove explicit recursion in retainer profiling (fixes #14758).

I've addressed all of the comments, please tell me if I need so cleanup anything else there.

Tue, Nov 20, 12:54 PM

Mon, Nov 19

mpickering awarded D5351: Remove explicit recursion in retainer profiling (fixes #14758) a Yellow Medal token.
Mon, Nov 19, 5:29 AM
qnikst updated the diff for D5351: Remove explicit recursion in retainer profiling (fixes #14758).
  • Address review comments.
Mon, Nov 19, 5:17 AM

Nov 18 2018

qnikst added a comment to D5351: Remove explicit recursion in retainer profiling (fixes #14758).
In D5351#147266, @osa1 wrote:

I just remembered the full command I was using as a test for this bug: ghc-stage2 --interactive +RTS -hr <<(echo 'sequence_ (replicate 1000000 (return ()))')

Nov 18 2018, 8:04 AM
qnikst added a comment to D5351: Remove explicit recursion in retainer profiling (fixes #14758).
In D5351#147264, @osa1 wrote:

Thanks! One thing that reliably triggered this problem in the past (at least on my system) was running profiled GHCi (build profiled GHC using the prof flavor, then run ghc-stage2 --interactive +RTS -hr). Could you give it a try without this patch to see if it still fails, and then try with this patch?

Nov 18 2018, 8:00 AM
qnikst created D5351: Remove explicit recursion in retainer profiling (fixes #14758).
Nov 18 2018, 7:37 AM

Dec 13 2016

qnikst updated the diff for D2827: Fix testcase T12903..

Try to remove haddock update.

Dec 13 2016, 3:07 PM
qnikst updated the diff for D2827: Fix testcase T12903..

Revert hadock update.

Dec 13 2016, 3:05 PM
qnikst updated the diff for D2827: Fix testcase T12903..

Revert disabling test on macos.

Dec 13 2016, 3:03 PM

Dec 11 2016

qnikst retitled D2827: Fix testcase T12903. from to Fix testcase T12903..
Dec 11 2016, 3:52 PM
qnikst added a comment to D2821: Disable T12903 due to flakiness.
diff --git a/testsuite/tests/rts/T12903.hs b/testsuite/tests/rts/T12903.hs
index ddaf8b9..f17bfa1 100644
--- a/testsuite/tests/rts/T12903.hs
+++ b/testsuite/tests/rts/T12903.hs
@@ -1,10 +1,20 @@
 import Control.Concurrent
 import Control.Exception
+import System.IO
 import System.Posix
+import System.Posix.IO
Dec 11 2016, 12:24 PM
qnikst added a comment to D2821: Disable T12903 due to flakiness.

I think i can just fix test to be more reliable without magic timeouts. Will it work?

Dec 11 2016, 4:29 AM

Dec 1 2016

qnikst updated the diff for D2770: Install toplevel handler inside fork..

Address review comments.

Dec 1 2016, 12:59 PM

Nov 30 2016

qnikst updated the diff for D2770: Install toplevel handler inside fork..

Remove debug code.

Nov 30 2016, 8:12 AM
qnikst retitled D2770: Install toplevel handler inside fork. from to Install toplevel handler inside fork..
Nov 30 2016, 8:02 AM

Nov 18 2016

qnikst retitled D2729: Allow to unregister threadWaitReadSTM action. from to Allow to unregister threadWaitReadSTM action..
Nov 18 2016, 1:56 PM

Sep 24 2015

qnikst updated subscribers of D1264: Merge Compacts into GHC.
Sep 24 2015, 3:15 PM

Jul 1 2015

qnikst updated subscribers of D1024: fix EBADF unqueueing in select backend (Trac #10590).
Jul 1 2015, 3:34 AM

Mar 4 2015

qnikst updated the diff for D637: Improve core linter so it catches unsafeCoerce problems (T9122).
  • Address review warnings
Mar 4 2015, 3:28 PM

Feb 17 2015

qnikst added a comment to D637: Improve core linter so it catches unsafeCoerce problems (T9122).
In D637#18657, @simonpj wrote:

I don't think we should allow any coercions between SIMD vectors, at least not yet. It's very architecture specific. I've added a note to the wiki page

Simon

Feb 17 2015, 3:44 PM
qnikst updated the diff for D637: Improve core linter so it catches unsafeCoerce problems (T9122).
  • Implement rules for vector coercion.
  • Specification update.
Feb 17 2015, 3:41 PM

Feb 14 2015

qnikst added a comment to D637: Improve core linter so it catches unsafeCoerce problems (T9122).

I have addressed all comments except:

Feb 14 2015, 6:15 PM
qnikst updated the diff for D637: Improve core linter so it catches unsafeCoerce problems (T9122).
  • don't check phantom types
Feb 14 2015, 6:10 PM
qnikst updated the diff for D637: Improve core linter so it catches unsafeCoerce problems (T9122).
  • Update comments
  • drop redundant check
  • Improve performance
  • Rewrite link check according to semantics
  • Update link in specification
  • Spec update
Feb 14 2015, 5:13 PM

Feb 12 2015

qnikst added a comment to D637: Improve core linter so it catches unsafeCoerce problems (T9122).

I have updated code according to wiki page. One thing is still missing - an update to coreSpec.pdf because I'm not sure how express that rule better.

Feb 12 2015, 4:29 AM
qnikst updated the diff for D637: Improve core linter so it catches unsafeCoerce problems (T9122).
  • Revert primRepActiveSizeB
  • Output warning instead of error on corelint
  • During unsafeCoerce check in linter, forbid unboxed tuples of the different sizes
  • Better support for unboxed tuples
  • Port tests to the new API
Feb 12 2015, 4:25 AM

Feb 10 2015

qnikst added a comment to D637: Improve core linter so it catches unsafeCoerce problems (T9122).
In D637#18348, @simonpj wrote:

Moreover, the wiki page is getting confused by conversations. Could you

Feb 10 2015, 10:10 AM
qnikst added a comment to D637: Improve core linter so it catches unsafeCoerce problems (T9122).
In D637#18348, @simonpj wrote:

I have attempted to respond, but it's hard for me to be sure I understand your questions.

Feb 10 2015, 5:48 AM

Feb 9 2015

qnikst added a comment to D637: Improve core linter so it catches unsafeCoerce problems (T9122).
In D637#18259, @simonpj wrote:

I've added responses to the wiki page

Feb 9 2015, 3:07 PM

Feb 3 2015

qnikst added a comment to D637: Improve core linter so it catches unsafeCoerce problems (T9122).
In D637#18188, @simonpj wrote:

Thanks. I have responded to questions on the wiki page, and asked some of my own

Feb 3 2015, 3:37 PM

Feb 2 2015

qnikst added a comment to D637: Improve core linter so it catches unsafeCoerce problems (T9122).
In D637#18182, @simonpj wrote:

I think it would be helpful to write specification of which this patch is the implementation.

That means writing a GHC Trac wiki page, describing the motivation, and listing precisely what inconsistencies caught by the new checks.

Then we can agree the spec. Then we can see how well the code implements it. I don't really feel able to respond to the details above until I know exactly what we are trying to achieve.

OK?

Thanks

Simon

Feb 2 2015, 5:18 PM

Jan 30 2015

qnikst updated the diff for D637: Improve core linter so it catches unsafeCoerce problems (T9122).
  • fixup! Add support for coercion check for UnboxedTuples.
  • Fix CoreLint file.
  • fixup! First attempt to update specification
Jan 30 2015, 1:47 PM

Jan 29 2015

qnikst added a comment to D637: Improve core linter so it catches unsafeCoerce problems (T9122).

I don't see why an unboxed tuple needs to be a "firstlevel citizen" in TyCon. I assume "firstlevel citizen" means a separate constructor, like UnboxedTupleTyCon. Instead of writing primRepActiveSizeB in TyCon and working over PrimRep, it could just be written in the Type module and accept a Type, or perhaps a RepType. Then, you can just sum the sizes of the unboxed tuple components, if that's the semantics that we decide on.

Jan 29 2015, 10:04 AM
qnikst updated the diff for D637: Improve core linter so it catches unsafeCoerce problems (T9122).
  • fixup! Improve core linter so it catches unsafeCoerce problems (T9122)
  • Add support for coercion check for UnboxedTuples.
  • First attempt to update specification
Jan 29 2015, 10:00 AM
qnikst added a comment to D637: Improve core linter so it catches unsafeCoerce problems (T9122).

About your tasks to do: what's blocking you? How can we help?

Jan 29 2015, 7:06 AM

Jan 28 2015

qnikst updated the diff for D637: Improve core linter so it catches unsafeCoerce problems (T9122).
  • Added note about bad coercions
Jan 28 2015, 12:43 PM
qnikst added a comment to D637: Improve core linter so it catches unsafeCoerce problems (T9122).

This is a diff for an early review, as some more changes are required:

Jan 28 2015, 8:37 AM
qnikst retitled D637: Improve core linter so it catches unsafeCoerce problems (T9122) from to Improve core linter so it catches unsafeCoerce problems (T9122).
Jan 28 2015, 8:30 AM

Jan 24 2015

qnikst added a comment to D626: Drop transformers dependency from ghc package..
In D626#17944, @hvr wrote:

Something like that was done with the binary package: ghc-binary was bundled prior to GHC 7.2, but starting with GHC 7.2 the real binary was bundled with GHC instead as there were problems with that approach (see Trac #5027)

Jan 24 2015, 2:00 AM

Jan 23 2015

qnikst added a comment to D626: Drop transformers dependency from ghc package..

There is another way in freezing libraries that 'ghc library' depends on by introducing ghc-transformers or alike. Then ghc-tf-instances can be a wrapper for real world tf and ghc bundled one.
The reason why this patch may be relevant for distro developers (you can see attached thread) is that this dependency really blocks user from installing another tf version and any library that depends on ghc library in the same environment. And in the most distributives (nixos is one of the notable exceptions) there is one global environment for the libraries.
I'm don't insist on applying this solution but some improvement on the front will be really helpful.

Jan 23 2015, 2:18 PM

Jan 22 2015

qnikst retitled D627: Make -ddump-splices output to stdout (fixes #8796) from to Make -ddump-splices output to stdout (fixes #8796).
Jan 22 2015, 3:02 PM
qnikst added a comment to D626: Drop transformers dependency from ghc package..

This Diff is done due to the transformers removal proposal on ghc-devs mailing list:

Jan 22 2015, 6:36 AM
qnikst retitled D626: Drop transformers dependency from ghc package. from to Drop transformers dependency from ghc package..
Jan 22 2015, 6:32 AM

Dec 25 2014

qnikst updated the diff for D592: Fix T9384: fix increacing capabilites number for eventlog..
  1. Indeintation fixed;
  2. Added comment why from=0 case is special.
Dec 25 2014, 5:19 PM
qnikst retitled D592: Fix T9384: fix increacing capabilites number for eventlog. from to Fix T9384: fix increacing capabilites number for eventlog..
Dec 25 2014, 5:10 PM

Dec 3 2014

qnikst updated subscribers of D550: Implement -XStaticValues.
Dec 3 2014, 3:11 PM