1323226 - Bug 1018486 disabled a major CC optimization on opt builds

Status: RESOLVED FIXED

(Core :: XPConnect, defect)

Description

[Olli Pettay [:smaug][[e-mail]]]

http://searchfox.org/mozilla-central/rev/594937fec2e2fc45fa9308ba2fb964816631f017/js/xpconnect/src/nsXPConnect.cpp#344,349

I don't know if clang plugin now complains if I remove that #ifdef

Comment 1

[Olli Pettay [:smaug][[e-mail]]]

[Tracking Requested - why for this release]:
Higher cycle collection times are bad for responsiveness

Comment 2

[Andrew McCreight [:mccr8]]

It would be good to add a comment to the method so people can tell that it is a "load-bearing poster". https://www.youtube.com/watch?v=QRVExJZKIT8

Comment 3

[Olli Pettay [:smaug][[e-mail]]]

yeah, I'm adding a comment that the QI has side effects.

FWIW, the regression is 4x more GCed objects in CC graphs (median)

Comment 4

[Olli Pettay [:smaug][[e-mail]]]

Attachment: enable_unmark_gray.diff

Comment 5

[(no longer active)]

Comment on attachment 8818317 [details] [diff] [review]
enable_unmark_gray.diff

Review of attachment 8818317 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/xpconnect/src/nsXPConnect.cpp
@@ +345,2 @@
>      nsCOMPtr<nsIXPConnectWrappedJSUnmarkGray> wjsug =
>        do_QueryInterface(aWrappedJS);

My guess as to why mystor did this was compiler warnings.  Do you mind making wjsug DebugOnly<>?

Comment 6

[Olli Pettay [:smaug][[e-mail]]]

How would that now cause a warning when it wasn't before.
https://bugzilla.mozilla.org/show_bug.cgi?id=1279746#c10

Comment 7

[Andrew Sutherland [:asuth] (he/him)]

(In reply to Olli Pettay [:smaug] from comment #3)
> yeah, I'm adding a comment that the QI has side effects.

The comment in the patch right now is "QIing to nsIXPConnectWrappedJSUnmarkGray may have side effects!" which seems more mysterious than it needs to be.  "This QI unmarks gray as a side-effect which avoids large amounts of wasted addref/release pairs." seems to be more in-line with the intent.

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/8574c820840b
re-enable the accidentally disabled xpc_TryUnmarkWrappedGrayObject also in opt builds, r=ehsan

Comment 9

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

Backed out for static analysis bustage:

https://hg.mozilla.org/integration/mozilla-inbound/rev/8529b942c88344d0f62c40512007cc46878d1b89

Push with failures: https://treeherder.mozilla.org/#/jobs?repo=mozilla-inbound&revision=8574c820840b444f3c0040ac67c7f9edae24c110
Failure log: https://treeherder.mozilla.org/logviewer.html#?job_id=40535328&repo=mozilla-inbound

[task 2016-12-13T19:10:49.773115Z] 19:10:49     INFO -  /home/worker/workspace/build/src/sccache2/sccache /home/worker/workspace/build/src/clang/bin/clang -std=gnu99 -o derprint.o -c  -DNDEBUG -DTRIMMED=1 -DNSPR20 -DNSS_NO_INIT_SUPPORT -DUSE_UTIL_DIRECTLY -DNO_NSPR_10_SUPPORT -DSSL_DISABLE_DEPRECATED_CIPHER_SUITE_NAMES -DLINUX2_1 -DLINUX -Dlinux -DHAVE_STRERROR -DXP_UNIX -D_REENTRANT -DNSS_DISABLE_LIBPKIX -I/home/worker/workspace/build/src/security/nss/cmd/lib -I/home/worker/workspace/build/src/obj-firefox/security/nss/cmd/lib/lib_sectool -I/home/worker/workspace/build/src/obj-firefox/dist/include/nspr -I/home/worker/workspace/build/src/obj-firefox/dist/include/private/nss -I/home/worker/workspace/build/src/obj-firefox/dist/include/nss -I/home/worker/workspace/build/src/obj-firefox/dist/include    -fPIC  -include /home/worker/workspace/build/src/obj-firefox/mozilla-config.h -DMOZILLA_CLIENT -MD -MP -MF .deps/derprint.o.pp -Qunused-arguments  -Qunused-arguments -Wall -Wempty-body -Wignored-qualifiers -Wpointer-arith -Wsign-compare -Wtype-limits -Wunreachable-code -Wclass-varargs -Wloop-analysis -Werror=non-literal-null-conversion -Wstring-conversion -Wthread-safety -Wno-error=deprecated-declarations -Wno-error=array-bounds -fno-strict-aliasing -ffunction-sections -fdata-sections -fno-math-errno -pthread -pipe  -g -Xclang -load -Xclang /home/worker/workspace/build/src/obj-firefox/build/clang-plugin/libclang-plugin.so -Xclang -add-plugin -Xclang moz-check -Os -fno-omit-frame-pointer    /home/worker/workspace/build/src/security/nss/cmd/lib/derprint.c
[task 2016-12-13T19:10:49.829195Z] 19:10:49     INFO -  In file included from /home/worker/workspace/build/src/obj-firefox/js/xpconnect/src/Unified_cpp_js_xpconnect_src1.cpp:110:
[task 2016-12-13T19:10:49.834285Z] 19:10:49     INFO -  /home/worker/workspace/build/src/js/xpconnect/src/nsXPConnect.cpp:345:5: error: Unused "kungFuDeathGrip" 'nsCOMPtr<nsIXPConnectWrappedJSUnmarkGray>' objects constructed from temporary values are prohibited
[task 2016-12-13T19:10:49.834375Z] 19:10:49     INFO -      nsCOMPtr<nsIXPConnectWrappedJSUnmarkGray> wjsug =
[task 2016-12-13T19:10:49.834405Z] 19:10:49     INFO -      ^
[task 2016-12-13T19:10:49.834501Z] 19:10:49     INFO -  /home/worker/workspace/build/src/js/xpconnect/src/nsXPConnect.cpp:346:7: note: Please switch all accesses to this value to go through 'wjsug', or explicitly pass 'wjsug' to `mozilla::Unused`
[task 2016-12-13T19:10:49.834542Z] 19:10:49     INFO -        do_QueryInterface(aWrappedJS);
[task 2016-12-13T19:10:49.834568Z] 19:10:49     INFO -        ^
[task 2016-12-13T19:10:49.834597Z] 19:10:49     INFO -  1 error generated.

Comment 10

[Olli Pettay [:smaug][[e-mail]]]

I'll add some Unused << and push to try, since I can't figure out how to run static analysis locally on Fedora.

Comment 11

[Olli Pettay [:smaug][[e-mail]]]

Attachment: enable_unmark_gray_with_unused.diff

https://treeherder.mozilla.org/#/jobs?repo=try&revision=84dae073d3b2d7a8d65c9e517f29d52137c247bc

Comment 12

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/6f6c33db62e1
re-enable the accidentally disabled xpc_TryUnmarkWrappedGrayObject also in opt builds, r=ehsan

Comment 13

[Phil Ringnalda (:philor)]

https://hg.mozilla.org/mozilla-central/rev/6f6c33db62e1

Comment 14

[Olli Pettay [:smaug][[e-mail]]]

Comment on attachment 8818337 [details] [diff] [review]
enable_unmark_gray_with_unused.diff

Approval Request Comment
[Feature/Bug causing the regression]: bug 1018486
[User impact if declined]: Longer CC times
[Is this code covered by automated tests?]: This code runs all the time, and telemetry shows there was regression
[Has the fix been verified in Nightly?]: Just landed to m-c
[Needs manual test from QE? If yes, steps to reproduce]:  No need. telemetry is enough here.
[List of other uplifts needed for the feature/fix]: NA
[Is the change risky?]: Not risky. Bringing back the old behavior
[Why is the change risky/not risky?]: Bringing back the old behavior.
[String changes made/needed]: NA

Comment 15

[Julien Cristau [:jcristau]]

Comment on attachment 8818337 [details] [diff] [review]
enable_unmark_gray_with_unused.diff

take cycle collection regression fix for aurora52

Comment 16

[Gerry Chang [:gchang]]

Comment on attachment 8818337 [details] [diff] [review]
enable_unmark_gray_with_unused.diff

Fix a higher cycle collection. Beta51+. Should be in 51 beta 8.

Comment 17

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/35f19f6baa6c

Comment 18

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/releases/mozilla-beta/rev/5aec4a29e94f

Comment 19

[Gerry Chang [:gchang]]

Track 51+ as regression.

Comment 20

[Ryan VanderMeulen [:RyanVM]]

Should we be considering this for esr45 as well?

Comment 21

[Olli Pettay [:smaug][[e-mail]]]

The code in esr45 is fine
https://dxr.mozilla.org/mozilla-esr45/source/js/xpconnect/src/nsXPConnect.cpp#310
since Bug 1279746 landed to FF50.