150015 - trace-malloc slows down builds even when not used

Status: RESOLVED FIXED

(Core :: XPCOM, defect, P1)

Description

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

trace-malloc slows down builds even when it's not used (no --trace-malloc
argument on the command line), by roughly a factor of 4, even though, as far as
I know, there's no way to get useful output from it (I might be wrong here,
though).  suppress_tracing is never even set to non-zero.  I'm going to attach a
patch that I think should fix this problem (and which makes my debug build
usable again).

Comment 1

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

Attachment: patch

Are there any ways of using trace-malloc that this would break?

Comment 2

[Brendan Eich [:brendan]]

Yes, this will break scripted uses (see dom/src/base/nsJSEnvironment.cpp, look
for TraceMalloc).

/be

Comment 3

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

I could never get those to work without using --trace-malloc on the command-line
anyway.  I thought it had something to do with an initial buffer size.  Maybe my
memory is wrong, though...

Comment 4

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

Attachment: patch

Pretty much the same thing, but fixes a bogus |return;|.  (For reference.  No
intent to check in.)

Comment 5

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

Attachment: patch

Here's the patch again (rewritten from the last one, since it uses different variable names).

Comment 6

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

Hrm, except there's comment 2.

Comment 7

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

(Except I'm not sure how the enable/disable functions ever worked dynamically, or what they'd be used for.)

Comment 8

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

Though, actually, this patch fixes Enable/Disable by setting tracing_enabled, so maybe it's ok (or as ok as it's always been).

Comment 9

[Brendan Eich [:brendan]]

Comment on attachment 274855 [details] [diff] [review]
patch

>+    if (!PR_Initialized() || !tracing_enabled) {

Transpose these || terms for faster results, right?

With the same change elsewhere, r=me.

/be

Comment 10

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

Comment on attachment 274855 [details] [diff] [review]
patch

This is a performance fix for the trace-malloc debugging tool that makes it so that trace-malloc doesn't degrade performance significantly when it is compiled in but not used.  This will make it easier for other developers to build trace-malloc and debug leaks without suffering the degradation.  No risk to releases, which don't ship this code.

Comment 11

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 274855 [details] [diff] [review]
patch

a=bzbarsky

Comment 12

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

Fix checked in to trunk, 2007-08-10 14:24.