1105851 - security/manager/boot/src/RootCertificateTelemetryUtils.cpp:68:68: error: 'ArrayLength' was not declared in this scope (non-SPS, non-unified)

Status: RESOLVED FIXED

(Core :: Security: PSM, defect)

Description

[Jan Beich]

On platforms lacking SPS profiler (Linux/PPC, BSDs, Solaris) non-unified build is broken due to header bootlegging. Barely noticed thanks to bug 1063726 and complete brokeness of OpenBSD buildbot.

In file included from security/manager/boot/src/DataStorage.cpp:7:
In file included from security/manager/boot/src/DataStorage.h:10:
In file included from dist/include/mozilla/Monitor.h:10:
In file included from dist/include/mozilla/CondVar.h:16:
In file included from dist/include/GeckoProfiler.h:201:
In file included from dist/include/GeckoProfilerImpl.h:17:
In file included from dist/include/PseudoStack.h:9:
dist/include/mozilla/ArrayUtils.h:14:2: error: "bootleg test"

security/manager/boot/src/RootCertificateTelemetryUtils.cpp:68:46: error: use of undeclared
      identifier 'ArrayLength'
  if (mozilla::BinarySearchIf(ROOT_TABLE, 0, ArrayLength(ROOT_TABLE),
                                             ^
1 error generated.

http://buildbot.rhaalovely.net/builders/mozilla-aurora-freebsd-amd64/builds/483/steps/build/logs/stdio

Comment 1

[Jan Beich]

(In reply to Jan Beich from comment #0)
> In file included from security/manager/boot/src/DataStorage.cpp:7:
> In file included from security/manager/boot/src/DataStorage.h:10:
> In file included from dist/include/mozilla/Monitor.h:10:
> In file included from dist/include/mozilla/CondVar.h:16:

Doh, too lazy to force SPS error from Try. Actual order before GeckProfiler.h is the following:

In file included from security/manager/boot/src/RootCertificateTelemetryUtils.cpp:7:
In file included from security/manager/boot/src/RootCertificateTelemetryUtils.h:10:
In file included from dist/include/mozilla/Telemetry.h:11:
In file included from dist/include/mozilla/StartupTimeline.h:33:

> In file included from dist/include/GeckoProfiler.h:201:
> In file included from dist/include/GeckoProfilerImpl.h:17:
> In file included from dist/include/PseudoStack.h:9:
> dist/include/mozilla/ArrayUtils.h:14:2: error: "bootleg test"

Comment 2

[Jan Beich]

Attachment: fix

Explicitly include ArrayUtils.h for ArrayLength().

Comment 3

[Landry Breuil (:gaston)]

Comment on attachment 8529899 [details] [diff] [review]
fix

Hah interesting i had seen that seemingly random failure, but only on aurora, i dont remember seeing it on central... and the local fix i had in my mq is more or less att 8529899, so lgtm!

Comment 4

[J.C. Jones [:jcj] (he/they)]

Comment on attachment 8529899 [details] [diff] [review]
fix

Review of attachment 8529899 [details] [diff] [review]:
-----------------------------------------------------------------

Absolutely good to me, too. Thanks!

Comment 5

[Jan Beich]

Attachment: fix

Trying again to confirm if |git bz attach| eats trailing whitespace in context lines.

Comment 6

[Jan Beich]

Attachment: fix

It does, uploading via browser. Carrying over r+ from attachment 8529899 [details] [diff] [review] as no changes except for the following context line with trailing whitespace:

 namespace mozilla { namespace psm { 

central: https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=a23080ad2540
beta:    https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=62c5bc83e526

Comment 7

[Jan Beich]

Comment on attachment 8530632 [details] [diff] [review]
fix

Approval Request Comment
[Feature/regressing bug #]: bug 1054498 regression
[User impact if declined]: Broken build on release channels for non-SPS platforms unless --enable-unified-compilation but beware of bug 1105781.
[Describe test coverage new/current, TBPL]: see comment 6, soon inbound
[Risks and why]: Low, broken build at most.
[String/UUID change made/needed]:

Comment 8

[Landry Breuil (:gaston)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/f50c458f6f69

> [User impact if declined]: Broken build on release channels for non-SPS platforms unless --enable-unified-compilation but beware of bug 1105781.

Not sure of this, i've seen broken builds on this failure in aurora (when aurora was 35) with the default unified compilation behaviour.

Comment 9

[Jan Beich]

--enable-unified-compilation is only enabled on Nightly but not for long [1]. FreeBSD ports use it like -jN to reduce load for package cluster of building 5 different flavors of Gecko engine.

[1] https://www.mail-archive.com/[e-mail]/msg11513.html

Comment 10

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/f50c458f6f69

Comment 11

[Ryan VanderMeulen (PTO, back July 27)]

https://hg.mozilla.org/releases/mozilla-aurora/rev/7eedf2ff7ff6
https://hg.mozilla.org/releases/mozilla-beta/rev/5615256dff5d