modocache (Brian Gesiak)
User

Projects

User does not belong to any projects.

User Details

User Since
Feb 16 2014, 1:10 AM (223 w, 4 h)

Recent Activity

Tue, May 22

modocache updated the diff for D46628: [ELF] Add --strip-debug-non-line option.

Thanks again for the reviews. I removed the extraneous else, and reverted the string switch back to what I had originally.

Tue, May 22, 8:38 PM
modocache added a comment to D46628: [ELF] Add --strip-debug-non-line option.

Does gold really preserve .debug_info and .debug_abbrev? Generally .debug_info is by far the largest DWARF section and so the one you most likely want to remove.

Tue, May 22, 8:09 PM
modocache added a comment to D46628: [ELF] Add --strip-debug-non-line option.

Sorry for letting this languish a bit. I took some time to experiment. Long story short I think llvm-dwarfdump and lldb aren't able to verify this option at the moment.

Tue, May 22, 6:36 PM

Mon, May 21

modocache committed rL332878: [DebugInfo] Fix typo "DWARG" in test comment (NFC).
[DebugInfo] Fix typo "DWARG" in test comment (NFC)
Mon, May 21, 12:52 PM

Sat, May 19

modocache committed rL332812: Re-revert "[Option] Fix PR37006 prefix choice in findNearest".
Re-revert "[Option] Fix PR37006 prefix choice in findNearest"
Sat, May 19, 9:25 AM
modocache added a comment to rC332804: [Driver] Loosen test for LLVM findNearest.

Yes, unfortunately I haven’t been able to reproduce this locally, on any build environment I have access to. There’s a discussion about this in https://reviews.llvm.org/D46776, but I can’t figure out why the buildbots wouldn’t behave the same way. Passing ‘-triple x86_64-scei-ps4-ubuntu’ as suggested there doesn’t make a difference for me locally. And indeed the argument parsing occurs before the triple is taken into account anyway, so it makes sense to me that it wouldn’t make a difference.

Sat, May 19, 5:57 AM
modocache committed rL332806: [Driver] Temporarily remove test for LLVM findNearest.
[Driver] Temporarily remove test for LLVM findNearest
Sat, May 19, 5:47 AM
modocache committed rC332806: [Driver] Temporarily remove test for LLVM findNearest.
[Driver] Temporarily remove test for LLVM findNearest
Sat, May 19, 5:47 AM
modocache committed rL332805: Un-revert "[Option] Fix PR37006 prefix choice in findNearest".
Un-revert "[Option] Fix PR37006 prefix choice in findNearest"
Sat, May 19, 5:07 AM
modocache committed rC332804: [Driver] Loosen test for LLVM findNearest.
[Driver] Loosen test for LLVM findNearest
Sat, May 19, 4:51 AM
modocache committed rL332804: [Driver] Loosen test for LLVM findNearest.
[Driver] Loosen test for LLVM findNearest
Sat, May 19, 4:51 AM

Tue, May 15

modocache added a comment to D46776: [Option] Fix PR37006 prefix choice in findNearest.

Hmm... but the >1 condition is such that an edit distance of greater than 1 results in no suggestion, whereas an edit distance of 1 or 0 results in a suggestion. In this case -- and apparently only on certain platforms' buildbots -- the edit distance between -debug-info-macros and -debug-info-macro is 1, and so it should display a suggestion.

Tue, May 15, 2:29 PM

Mon, May 14

modocache added a comment to D46776: [Option] Fix PR37006 prefix choice in findNearest.

I'm not yet sure why, but this change caused buildbot failures due to a test in Clang for the "did you mean" feature. I couldn't figure out why the buildbots would fail whereas the Clang tests pass for me locally, and so reverted this change for the time being, in rL332304. I'll try to resubmit once I've figured out and fixed the issue.

Mon, May 14, 3:43 PM
modocache committed rL332304: Revert "[Option] Fix PR37006 prefix choice in findNearest".
Revert "[Option] Fix PR37006 prefix choice in findNearest"
Mon, May 14, 3:40 PM
modocache added a comment to D46776: [Option] Fix PR37006 prefix choice in findNearest.

Thanks for the reviews, and for pointing out the bug!

Mon, May 14, 2:40 PM
modocache committed rL332299: [Option] Fix PR37006 prefix choice in findNearest.
[Option] Fix PR37006 prefix choice in findNearest
Mon, May 14, 2:38 PM
modocache closed D46776: [Option] Fix PR37006 prefix choice in findNearest.
Mon, May 14, 2:38 PM
modocache added inline comments to D46776: [Option] Fix PR37006 prefix choice in findNearest.
Mon, May 14, 9:36 AM

Fri, May 11

modocache created D46776: [Option] Fix PR37006 prefix choice in findNearest.
Fri, May 11, 2:37 PM

Thu, May 10

modocache added a comment to D43242: [Coroutines] PR34897: Fix incorrect elisions.

Hooray! Thanks again for the review, @GorNishanov.

Thu, May 10, 8:26 PM
modocache committed rL332077: [Coroutines] PR34897: Fix incorrect elisions.
[Coroutines] PR34897: Fix incorrect elisions
Thu, May 10, 8:16 PM
modocache closed D43242: [Coroutines] PR34897: Fix incorrect elisions.
Thu, May 10, 8:16 PM
modocache committed rL332075: [Support] Add docs for 'openFileFor{Write,Read}'.
[Support] Add docs for 'openFileFor{Write,Read}'
Thu, May 10, 6:51 PM
modocache closed D46499: [Support] Add docs for 'openFileFor{Write,Read}'.
Thu, May 10, 6:51 PM
modocache updated the diff for D46499: [Support] Add docs for 'openFileFor{Write,Read}'.

I updated this diff to not only document how the ResultFD is set, but also make it a little more consistent across platforms. Previously Windows would only sometimes set it to -1 in the case of failure, now it always does.

Thu, May 10, 2:17 PM
modocache added a comment to D46628: [ELF] Add --strip-debug-non-line option.

Please just bring back the first patch to line 1316 and submit. Thanks!

Thu, May 10, 11:39 AM
modocache added a comment to D46628: [ELF] Add --strip-debug-non-line option.

Thanks for all the review!

Thu, May 10, 10:21 AM

Wed, May 9

modocache updated the diff for D46628: [ELF] Add --strip-debug-non-line option.

Thanks for the review! I adopted the changes you suggested. As for validating the debug info, what about using 'llvm-dwarfdump -verify'? In fact running that on the output of this new option resulted in errors, so I'll address those.

Wed, May 9, 10:04 PM
modocache added a comment to D46499: [Support] Add docs for 'openFileFor{Write,Read}'.

Correct! I'll mention that as well.

Wed, May 9, 8:58 AM
modocache added a comment to D46499: [Support] Add docs for 'openFileFor{Write,Read}'.

Thanks for the comments!

Wed, May 9, 8:55 AM

Tue, May 8

modocache created D46628: [ELF] Add --strip-debug-non-line option.
Tue, May 8, 10:16 PM

Mon, May 7

modocache updated the diff for D43242: [Coroutines] PR34897: Fix incorrect elisions.

Sorry it took me a while to get around to this, but I think this new approach fits more along the lines of what you were thinking, @GorNishanov. I also confirmed that it optimizes the program you linked to just as well as before.

Mon, May 7, 9:46 AM

Sat, May 5

modocache created D46499: [Support] Add docs for 'openFileFor{Write,Read}'.
Sat, May 5, 9:11 PM

Fri, May 4

modocache added a comment to D45860: [Coroutines] Catch exceptions in await_resume.

Thanks again for all the reviews, @GorNishanov! Very much appreciated.

Fri, May 4, 7:09 AM
modocache committed rL331519: [Coroutines] Catch exceptions in await_resume.
[Coroutines] Catch exceptions in await_resume
Fri, May 4, 7:08 AM
modocache committed rC331519: [Coroutines] Catch exceptions in await_resume.
[Coroutines] Catch exceptions in await_resume
Fri, May 4, 7:07 AM
modocache closed D45860: [Coroutines] Catch exceptions in await_resume.
Fri, May 4, 7:07 AM

Wed, May 2

modocache updated the diff for D45860: [Coroutines] Catch exceptions in await_resume.

Oops, thanks for testing on release mode, @GorNishanov. Turns out I had a dangling pointer. With this update the tests pass on both release and debug.

Wed, May 2, 6:06 PM
modocache updated the diff for D45860: [Coroutines] Catch exceptions in await_resume.

Thanks for the review, @GorNishanov. Here's a more correct solution: an i1 is used to keep track of whether await_resume threw. If it did, the coroutine body is skipped, and we go straight to the final suspend point. Otherwise, the coroutine body is executed as normal.

Wed, May 2, 11:24 AM

Apr 21 2018

modocache committed rL330530: Add tests for llvm-bcanalyzer stream types.
Add tests for llvm-bcanalyzer stream types
Apr 21 2018, 5:08 PM
modocache committed rC330530: Add tests for llvm-bcanalyzer stream types.
Add tests for llvm-bcanalyzer stream types
Apr 21 2018, 5:07 PM
modocache closed D41980: Add tests for llvm-bcanalyzer stream types.
Apr 21 2018, 5:07 PM
modocache added a comment to D41979: [bcanalyzer] Recognize more stream types.

Strange, the revision appeared to have been "Accepted" by @JDevlieghere. I wonder why the message says it wasn't accepted...? Apologies in advance if I did something incorrectly when landing this.

Apr 21 2018, 5:00 PM
modocache committed rL330529: [bcanalyzer] Recognize more stream types.
[bcanalyzer] Recognize more stream types
Apr 21 2018, 4:56 PM
This revision was not accepted when it landed; it landed in state Needs Review.
Apr 21 2018, 4:56 PM
modocache updated the diff for D41979: [bcanalyzer] Recognize more stream types.

Thanks, @JDevlieghere! I added two tests like the ones you described. Thankfully it doesn't crash because llvm-bcanalyzer verifies the bitcode invariant that input sizes be a multiple of 4.

Apr 21 2018, 4:35 PM
modocache added a dependency for D41980: Add tests for llvm-bcanalyzer stream types: D41979: [bcanalyzer] Recognize more stream types.
Apr 21 2018, 10:12 AM
modocache added a dependent revision for D41979: [bcanalyzer] Recognize more stream types: D41980: Add tests for llvm-bcanalyzer stream types.
Apr 21 2018, 10:12 AM
modocache updated the diff for D41980: Add tests for llvm-bcanalyzer stream types.

Added -fmodule-format=raw.

Apr 21 2018, 10:11 AM
modocache updated the diff for D41979: [bcanalyzer] Recognize more stream types.

Sorry for letting this diff languish for so long. I updated the tests to be simple files with just the bytes necessary to be picked up as a specific stream type. I also moved the magic number reading over to a static helper function. I considered adding this function as a member of llvm::BirstreamReader, but chose not to for now, because doing so (as far as I can tell) would mean encoding Clang's magic numbers into LLVM's machinery. Let me know if you have different thoughts. Thanks!

Apr 21 2018, 9:50 AM

Apr 19 2018

modocache created D45860: [Coroutines] Catch exceptions in await_resume.
Apr 19 2018, 8:25 PM

Apr 12 2018

modocache committed rL329957: Let llvm-diff correctly deal with….
Let llvm-diff correctly deal with…
Apr 12 2018, 2:33 PM
modocache closed D34856: Let llvm-diff correctly deal with Undef/ConstantAggregateZero/ConstantVector/IndirectBr.
Apr 12 2018, 2:32 PM
modocache added a comment to D34856: Let llvm-diff correctly deal with Undef/ConstantAggregateZero/ConstantVector/IndirectBr.

Thanks for working on this, @aqjune! I was able to apply the patch cleanly, and all the tests build and passed, so I committed it as rL329957. Thanks again!

Apr 12 2018, 2:32 PM

Apr 5 2018

modocache committed rL329285: [Lexicon] Add "ICE", internal compiler error.
[Lexicon] Add "ICE", internal compiler error
Apr 5 2018, 7:13 AM

Apr 3 2018

modocache accepted D45221: [coroutines] Respect alloca alignment requirements when building coroutine frame.

Thank you, @GorNishanov, this is excellent! I've pulled this down and confirmed it fixes some nasty errors in a codebase I work on, related to this usage of std::aligned_storage. The problem also reproduced with this much smaller example: https://reviews.llvm.org/P8076 -- and I've confirmed this change of yours fixes that, too. So, thank you!

Apr 3 2018, 1:00 PM

Apr 2 2018

modocache committed rL329033: [Coroutines] Avoid assert splitting hidden coros.
[Coroutines] Avoid assert splitting hidden coros
Apr 2 2018, 4:42 PM
modocache closed D44185: [Coroutines] Avoid assert splitting hidden coros.
Apr 2 2018, 4:42 PM
modocache added reviewers for D44185: [Coroutines] Avoid assert splitting hidden coros: majnemer, rnk.

Oops! I used arc diff --verbatim and ended up removing reviewers.

Apr 2 2018, 3:43 PM
modocache updated the diff for D44185: [Coroutines] Avoid assert splitting hidden coros.

Thanks @rnk! I updated this as per your suggestions. As a result it's a smaller and less invasive change -- thank you! Let me know if this looks good.

Apr 2 2018, 2:20 PM
modocache added inline comments to D44185: [Coroutines] Avoid assert splitting hidden coros.
Apr 2 2018, 9:49 AM

Apr 1 2018

modocache added inline comments to D44185: [Coroutines] Avoid assert splitting hidden coros.
Apr 1 2018, 5:39 PM
modocache committed rL328951: [Coroutines] Schedule coro-split before asan.
[Coroutines] Schedule coro-split before asan
Apr 1 2018, 4:58 PM
modocache committed rC328951: [Coroutines] Schedule coro-split before asan.
[Coroutines] Schedule coro-split before asan
Apr 1 2018, 4:58 PM
modocache closed D43927: [Coroutines] Schedule coro-split before asan.
Apr 1 2018, 4:58 PM
modocache added a comment to D44552: [Coroutines] Find custom allocators in class scope.

Thanks for the review, @GorNishanov! I adopted your suggestions and made the commit.

Apr 1 2018, 4:05 PM
modocache committed rC328949: [Coroutines] Find custom allocators in class scope.
[Coroutines] Find custom allocators in class scope
Apr 1 2018, 4:04 PM
modocache committed rL328949: [Coroutines] Find custom allocators in class scope.
[Coroutines] Find custom allocators in class scope
Apr 1 2018, 4:04 PM
modocache closed D44552: [Coroutines] Find custom allocators in class scope.
Apr 1 2018, 4:04 PM
modocache accepted D45120: [coroutines] Add __builtin_coro_noop => llvm.coro.noop.

Great, thanks!

Apr 1 2018, 3:17 PM
modocache accepted D45114: [coroutines] Add support for llvm.coro.noop intrinsics.

Thanks!

Apr 1 2018, 3:14 PM

Mar 27 2018

modocache accepted D37115: [coroutines] Do not attempt to typo-correct when coroutine is looking for required members.

Excellent, thanks!

Mar 27 2018, 10:19 AM

Mar 26 2018

modocache added a comment to D44915: [coroutines] Fix invalid source range in co_await call expressions..

Excellent, thank you! I think this might fix PR34982?

Mar 26 2018, 8:26 PM

Mar 20 2018

modocache updated subscribers of D43219: [PartialInlining] Fix Crash from holding a reference to a destructed ORE.

Thanks for tracking down this bug! You may wish to ask @anemet for a review, I think he would have a good idea of what a proper fix would look like. As-is this LGTM, although in my opinion the test case could benefit from being reduced further.

Mar 20 2018, 5:53 AM

Mar 19 2018

modocache created D44672: [CodeGen] Disable UBSan for coroutine functions.
Mar 19 2018, 9:05 PM

Mar 15 2018

modocache created D44552: [Coroutines] Find custom allocators in class scope.
Mar 15 2018, 11:26 PM

Mar 7 2018

modocache updated the diff for D43242: [Coroutines] PR34897: Fix incorrect elisions.

Just tidying up one line with clang-format. I guess everyone's busy preparing for the Jacksonville standards meeting. Please give this a review when you all have a second -- in addition to the reported bug, https://bugs.llvm.org/show_bug.cgi?id=34897, I've also confirmed that this diff fixes ASAN errors that surfaced in a project at work. Thanks! :)

Mar 7 2018, 9:04 PM

Mar 6 2018

modocache updated the diff for D44185: [Coroutines] Avoid assert splitting hidden coros.

Update formatting with clang-format.

Mar 6 2018, 7:10 PM
modocache created D44185: [Coroutines] Avoid assert splitting hidden coros.
Mar 6 2018, 7:02 PM

Feb 28 2018

modocache added a comment to D43927: [Coroutines] Schedule coro-split before asan.

I wasn't sure what the best way to test this would be. The assertion occurs in LLVM, but Clang is responsible for scheduling the passes. If anyone has any suggestions, I'd greatly appreciate them!

Feb 28 2018, 8:37 PM
modocache created D43927: [Coroutines] Schedule coro-split before asan.
Feb 28 2018, 8:36 PM

Feb 22 2018

modocache added a comment to D43242: [Coroutines] PR34897: Fix incorrect elisions.

Friendly ping! Let me know if this seems totally wrong to you, @GorNishanov :)

Feb 22 2018, 7:17 AM

Feb 20 2018

modocache added inline comments to D43473: [mem2reg] Use range loops (NFCI).
Feb 20 2018, 5:12 AM

Feb 19 2018

modocache committed rL325539: Revert "[mem2reg] Use range loops (NFCI)".
Revert "[mem2reg] Use range loops (NFCI)"
Feb 19 2018, 2:51 PM
modocache added a comment to D43473: [mem2reg] Use range loops (NFCI).

Great, thank you both for the reviews!

Feb 19 2018, 1:56 PM
modocache committed rL325532: [mem2reg] Use range loops (NFCI).
[mem2reg] Use range loops (NFCI)
Feb 19 2018, 1:47 PM
modocache closed D43473: [mem2reg] Use range loops (NFCI).
Feb 19 2018, 1:47 PM
modocache committed rL325529: [Coroutines] Move debug statement before assert.
[Coroutines] Move debug statement before assert
Feb 19 2018, 12:53 PM
modocache retitled D43473: [mem2reg] Use range loops (NFCI) from Bump iterator in for loop increment expr (NFCI) to [mem2reg] Use range loops (NFCI).
Feb 19 2018, 10:38 AM
modocache updated the diff for D43473: [mem2reg] Use range loops (NFCI).

Thanks for the review! I've updated the diff to use range loops instead.

Feb 19 2018, 10:38 AM
modocache created D43473: [mem2reg] Use range loops (NFCI).
Feb 19 2018, 10:01 AM

Feb 16 2018

modocache committed rC325342: [Coroutines] Use target-agnostic size_t in test.
[Coroutines] Use target-agnostic size_t in test
Feb 16 2018, 6:15 AM
modocache committed rL325342: [Coroutines] Use target-agnostic size_t in test.
[Coroutines] Use target-agnostic size_t in test
Feb 16 2018, 6:15 AM

Feb 15 2018

modocache added a comment to D42606: [Coroutines] Use allocator overload when available.

Thank you for all the reviews, @GorNishanov!

Feb 15 2018, 12:47 PM
modocache committed rC325291: [Coroutines] Use allocator overload when available.
[Coroutines] Use allocator overload when available
Feb 15 2018, 12:41 PM
modocache committed rL325291: [Coroutines] Use allocator overload when available.
[Coroutines] Use allocator overload when available
Feb 15 2018, 12:41 PM
modocache closed D42606: [Coroutines] Use allocator overload when available.
Feb 15 2018, 12:41 PM
modocache added a comment to D42605: [Sema] Toggle diags when finding allocators (NFCI).

Thanks again, @GorNishanov, for all the reviews!

Feb 15 2018, 12:14 PM
modocache committed rL325288: [Sema] Toggle diags when finding allocators (NFCI).
[Sema] Toggle diags when finding allocators (NFCI)
Feb 15 2018, 12:11 PM
modocache committed rC325288: [Sema] Toggle diags when finding allocators (NFCI).
[Sema] Toggle diags when finding allocators (NFCI)
Feb 15 2018, 12:11 PM