645423 - nsPresArena.cpp fails to build with -DDEBUG_TRACEMALLOC_PRESARENA. (ARENA_POISON undefined)

Status: RESOLVED FIXED

(Core :: Layout, defect)

Description

[Mats Palmgren (inactive)]

layout/base/nsPresArena.cpp: In static member function 'static PRUword nsPresArena::GetPoisonValue()':
layout/base/nsPresArena.cpp:471: error: 'ARENA_POISON' was not declared in this scope
make[2]: *** [nsPresArena.o] Error 1

Comment 1

[Mats Palmgren (inactive)]

Attachment: Always init poison value

Move the setup of the poison value outside the
#ifndef DEBUG_TRACEMALLOC_PRESARENA
so it's always available.

Call the one-time init function also when DEBUG_TRACEMALLOC_PRESARENA
is defined.

This fixes the compilation error.

Comment 2

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Comment on attachment 522165 [details] [diff] [review]
Always init poison value

r=dbaron.  Thanks for fixing this.

Comment 3

[Mats Palmgren (inactive)]

Attachment: Don't poison nsRuleData in --enable-valgrind build

Also, wouldn't it be better to leave 'mValueOffsets' uninitialized
in a --enable-valgrind build and let valgrind catch any memory errors?

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Comment on attachment 522167 [details] [diff] [review]
Don't poison nsRuleData in --enable-valgrind build

r=dbaron, except I think it's simpler to skip the .h change and just have an #ifndef MOZ_VALGRIND block inside the destructor just like the one in the constructor.  No point worrying about inlining in DEBUG builds, and it makes the code easier to understand.

Comment 5

[Mats Palmgren (inactive)]

Attachment: Don't poison nsRuleData in --enable-valgrind build

Ok, updated as suggested.

Comment 6

[Mats Palmgren (inactive)]

Fixed in Cedar:
http://hg.mozilla.org/projects/cedar/rev/7e0d493e7b10
http://hg.mozilla.org/projects/cedar/rev/81d1df28a908

Comment 7

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

http://hg.mozilla.org/mozilla-central/rev/7e0d493e7b10
http://hg.mozilla.org/mozilla-central/rev/81d1df28a908