This is an archive of the discontinued LLVM Phabricator instance.

Differential D73966

[analyzer] Add 10.0.0 release notes.
ClosedPublic

Authored by Szelethus on Feb 4 2020, 7:07 AM.

Details

Reviewers
hans
NoQ
Charusso
xazax.hun
baloghadamsoftware
balazske
gamesh411
rnkovacs
dcoughlin
Commits
rG64515b35844b: [analyzer] Add 10.0.0 release notes.
Summary

I was MIA for a couple months, so I'm very unsure whether I caught everything, but here is what happened since 9.0.0.

Diff Detail

Event Timeline

Szelethus created this revision.Feb 4 2020, 7:07 AM
Herald added subscribers: cfe-commits, phosek, dkrupp and 5 others. · View Herald TranscriptFeb 4 2020, 7:07 AM
Szelethus added a reviewer: dcoughlin.Feb 4 2020, 7:08 AM
Szelethus marked an inline comment as done.
Szelethus added inline comments.
clang/docs/ReleaseNotes.rst
418

I tried my best here but didn't get far. :)

xazax.hun added inline comments.Feb 4 2020, 8:15 AM
clang/docs/ReleaseNotes.rst
420

I'd just say Numerous smaller fixes. or Numerous other improvements.

NoQ added a comment.Feb 4 2020, 8:47 AM

Thanks!!

clang/docs/ReleaseNotes.rst
408

I think we don't traditionally highlight in release notes that something was a result of GSoC. I don't have an opinion here, just don't see much of it.

408

I think this one's out of alpha already (i.e., cplusplus.PlacementNew). I'd also bump it to the top because it applies to more users.

413

@Charusso's work was already mentioned in the previous release notes because it was already partially enabled mid-summer.

418

Not much here, mostly bugfixes.

NoQ added inline comments.Feb 4 2020, 10:15 AM
clang/docs/ReleaseNotes.rst
404–405

Let's make this more concise: "Dead stores checker now warns about nested dead stores.". While it was originally unclear whether this should be on by default, i think there's relatively little motivation in disabling this option now that everybody liked it :)

418

Oh, wait, there was this D71728 thing.

New checker: security.insecureAPI.decodeValueOfObjCType warns about potential buffer overflows when using [NSCoder decodeValueOfObjCType:at:].

(probably no need for a separate section for Objective-C)

Szelethus marked 2 inline comments as done.Feb 4 2020, 10:51 AM
Szelethus added inline comments.
clang/docs/ReleaseNotes.rst
408

Yup, but AFAIK it moved out of alpha after rc1 was tagged.

408

Yup, sure.

NoQ mentioned this in D74004: [analyzer] Move fuchsia.Lock checker to alpha.Feb 4 2020, 4:07 PM
balazske added inline comments.Feb 5 2020, 12:47 AM
clang/docs/ReleaseNotes.rst
408

alpha.cplusplus.PlacementNew not alpha.plusplus.PlacementNew?

Charusso added inline comments.Feb 5 2020, 12:55 AM
clang/docs/ReleaseNotes.rst
408

cplusplus.PlacementNew
(cf. https://github.com/llvm/llvm-project/commit/bc29069dc40)

Szelethus updated this revision to Diff 242539.Feb 5 2020, 2:37 AM
Szelethus marked 10 inline comments as done.
Szelethus retitled this revision from [analyzer][WIP] Add 10.0.0 release notes. to [analyzer] Add 10.0.0 release notes..
Szelethus added inline comments.
clang/docs/ReleaseNotes.rst
408

Nice catch ;) I checked out the llvmorg-10.0.0-rc1 tag and can confirm that at that point, the checker is still in alpha, unfortunately.

NoQ added inline comments.Feb 5 2020, 4:22 PM
clang/docs/ReleaseNotes.rst
405

D74004

  1. The checker is now in alpha.
  2. This checker wasn't enabled in the release either, as far as i understand.
xazax.hun added inline comments.Feb 5 2020, 4:59 PM
clang/docs/ReleaseNotes.rst
405

It was! This is about the HandleChecker the other is about the LockChecker.

Szelethus marked 3 inline comments as done.Feb 6 2020, 4:08 AM
Szelethus added inline comments.
clang/docs/ReleaseNotes.rst
405

Yup, this seems to be correct in the release notes.

Szelethus updated this revision to Diff 242860.Feb 6 2020, 4:10 AM
Szelethus marked an inline comment as done.

Actually update the revision.

Szelethus updated this revision to Diff 242869.Feb 6 2020, 4:22 AM

Actually, actually upload the correct one. Getting rusty eh.

hans accepted this revision.Feb 10 2020, 5:34 AM

Looks good to me. Thanks for writing release notes!

If you have commit access, go ahead and push directly to the branch, otherwise let me know and I'll be happy to do it for you.

This revision is now accepted and ready to land.Feb 10 2020, 5:34 AM
Herald added a subscriber: steakhal. · View Herald TranscriptFeb 10 2020, 5:34 AM
Szelethus added a comment.Feb 10 2020, 5:58 AM

I missed out on the transition to github, so I suspect that the commit access will only be extended to it after tagging rc2. I think it would be better if you committed this on my behalf, thanks! :)

hans added a comment.Feb 10 2020, 6:35 AM

Pushed to 10.x in 64515b35844b925bdb76821d03ad4d7ddebe06e7. Thanks!

Szelethus closed this revision.Feb 10 2020, 9:08 AM

Cheers!

Revision Contents

PathSize
clang/
docs/
22 lines

Diff 242869

clang/docs/ReleaseNotes.rst

Show First 20 Lines • Show All 395 Lines • ▼ Show 20 Lines
libclang libclang
-------- --------
- ... - ...
Static Analyzer Static Analyzer
--------------- ---------------
- The Clang analyzer checker ``DeadStores`` gets a new option called - New checker: ``alpha.cplusplus.PlacementNew`` to detect whether the storage
NoQUnsubmitted
Done
ReplyInline Actions

Let's make this more concise: "Dead stores checker now warns about nested dead stores.". While it was originally unclear whether this should be on by default, i think there's relatively little motivation in disabling this option now that everybody liked it :)

NoQ: Let's make this more concise: "Dead stores checker now warns about nested dead stores.". While…
NoQUnsubmitted
Done
ReplyInline Actions

D74004

  1. The checker is now in alpha.
  2. This checker wasn't enabled in the release either, as far as i understand.
NoQ: D74004 1) The checker is now in alpha. 2) This checker wasn't enabled in the release either…
xazax.hunUnsubmitted
Done
ReplyInline Actions

It was! This is about the HandleChecker the other is about the LockChecker.

xazax.hun: It was! This is about the HandleChecker the other is about the LockChecker.
SzelethusAuthorUnsubmitted
Done
ReplyInline Actions

Yup, this seems to be correct in the release notes.

Szelethus: Yup, this seems to be correct in the release notes.
``WarnForDeadNestedAssignments`` to detect nested dead assignments provided for default placement new is sufficiently large.
(enabled by default).
- ... - New checker: ``fuchsia.HandleChecker`` to detect leaks related to Fuchsia
NoQUnsubmitted
Done
ReplyInline Actions

I think we don't traditionally highlight in release notes that something was a result of GSoC. I don't have an opinion here, just don't see much of it.

NoQ: I think we don't traditionally highlight in release notes that something was a result of GSoC.
SzelethusAuthorUnsubmitted
Done
ReplyInline Actions

Yup, sure.

Szelethus: Yup, sure.
NoQUnsubmitted
Done
ReplyInline Actions

I think this one's out of alpha already (i.e., cplusplus.PlacementNew). I'd also bump it to the top because it applies to more users.

NoQ: I think this one's out of alpha already (i.e., `cplusplus.PlacementNew`). I'd also bump it to…
SzelethusAuthorUnsubmitted
Done
ReplyInline Actions

Yup, but AFAIK it moved out of alpha after rc1 was tagged.

Szelethus: Yup, but AFAIK it moved out of alpha after rc1 was tagged.
balazskeUnsubmitted
Done
ReplyInline Actions

alpha.cplusplus.PlacementNew not alpha.plusplus.PlacementNew?

balazske: `alpha.cplusplus.PlacementNew` not `alpha.plusplus.PlacementNew`?
CharussoUnsubmitted
Done
ReplyInline Actions

cplusplus.PlacementNew
(cf. https://github.com/llvm/llvm-project/commit/bc29069dc40)

Charusso: `cplusplus.PlacementNew` (cf. https://github.com/llvm/llvm-project/commit/bc29069dc40)
SzelethusAuthorUnsubmitted
Done
ReplyInline Actions

Nice catch ;) I checked out the llvmorg-10.0.0-rc1 tag and can confirm that at that point, the checker is still in alpha, unfortunately.

Szelethus: Nice catch ;) I checked out the `llvmorg-10.0.0-rc1` tag and can confirm that at that point…
handles.
- New checker: ``security.insecureAPI.decodeValueOfObjCType`` warns about
potential buffer overflows when using ``[NSCoder decodeValueOfObjCType:at:]``
NoQUnsubmitted
Done
ReplyInline Actions

@Charusso's work was already mentioned in the previous release notes because it was already partially enabled mid-summer.

NoQ: @Charusso's work was already mentioned in the [[ http://releases.llvm.org/9.0.
- ``deadcode.DeadStores`` now warns about nested dead stores.
- Condition values that are relevant to the occurance of a bug are far better
explained in bug reports.
SzelethusAuthorUnsubmitted
Done
ReplyInline Actions

I tried my best here but didn't get far. :)

Szelethus: I tried my best here but didn't get far. :)
NoQUnsubmitted
Done
ReplyInline Actions

Not much here, mostly bugfixes.

NoQ: Not much here, mostly bugfixes.
NoQUnsubmitted
Done
ReplyInline Actions

Oh, wait, there was this D71728 thing.

New checker: security.insecureAPI.decodeValueOfObjCType warns about potential buffer overflows when using [NSCoder decodeValueOfObjCType:at:].

(probably no need for a separate section for Objective-C)

NoQ: Oh, wait, there was this D71728 thing. > New checker: `security.insecureAPI.
- Despite still being at an alpha stage, checkers implementing taint analyses
and C++ iterator rules were improved greatly.
xazax.hunUnsubmitted
Done
ReplyInline Actions

I'd just say Numerous smaller fixes. or Numerous other improvements.

xazax.hun: I'd just say `Numerous smaller fixes.` or `Numerous other improvements.`
- Numerous smaller fixes.
.. _release-notes-ubsan: .. _release-notes-ubsan:
Undefined Behavior Sanitizer (UBSan) Undefined Behavior Sanitizer (UBSan)
------------------------------------ ------------------------------------
- * The ``pointer-overflow`` check was extended added to catch the cases where - * The ``pointer-overflow`` check was extended added to catch the cases where
a non-zero offset is applied to a null pointer, or the result of a non-zero offset is applied to a null pointer, or the result of
▲ Show 20 Lines • Show All 65 Lines • Show Last 20 Lines