1320196 - Eliminate unnecessary string concatenations in ImageCacheKey::ComputeHash

Status: RESOLVED FIXED

(Core :: Graphics: ImageLib, defect, P3)

Description

[Andrew Osmond [:aosmond] (he/him)]

Hashing several strings appended together, and hashing them individually should result in the same value, except the latter is a much cheaper option.

http://searchfox.org/mozilla-central/rev/b4e6fa3527436bdbcd9dd2e1982382fe423431f3/image/ImageCacheKey.cpp#158

This allow triggers a malloc, which shows up in crash signatures, such as:

https://crash-stats.mozilla.com/report/index/77317c14-819d-4853-99ac-7bcaf2161118

The malloc is so big (presumably) because the URI spec can be quite large if it is just a base64 encoded image instead of a normal URL.

Comment 1

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Remove string concat in ImageCacheKey::ComputeHash, v1

try: https://treeherder.mozilla.org/#/jobs?repo=try&revision=dec4057d18498db22bed7480dd730c991746da70

Comment 2

[Timothy Nikkel (:tnikkel)]

Comment on attachment 8814245 [details] [diff] [review]
Remove string concat in ImageCacheKey::ComputeHash, v1

Nice!

ImageURL::ComputeHash isn't a generic function, but rather specific to ImageCacheKey. So maybe make it private and mark ImageCacheKey as a friend class?

Comment 3

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Remove string concat in ImageCacheKey::ComputeHash, v2 [carries r=tnikkel]

Incorporate review feedback.

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/b8cdb59ac18b
Avoid string concatenation when computing the hash for ImageCacheKey. r=tnikkel

Comment 5

[Phil Ringnalda (:philor)]

https://hg.mozilla.org/mozilla-central/rev/b8cdb59ac18b

Comment 6

[Andrew Osmond [:aosmond] (he/him)]

Comment on attachment 8814389 [details] [diff] [review]
Remove string concat in ImageCacheKey::ComputeHash, v2 [carries r=tnikkel]

Approval Request Comment
[Feature/regressing bug #]: N/A
[User impact if declined]: May crash in low memory conditions.
[Describe test coverage new/current, TreeHerder]: The crash itself has only been observed in crash reports, not easily reproducible. However the crash occurs on a success path used every time an image is loaded and is thus well exercised by existing mochi/web-platorm/reftests.
[Risks and why]: Minimal, there was no functional change, only the need for some temporary allocations on the heap was removed.
[String/UUID change made/needed]: N/A

Comment 7

[Andrew Osmond [:aosmond] (he/him)]

Comment on attachment 8814389 [details] [diff] [review]
Remove string concat in ImageCacheKey::ComputeHash, v2 [carries r=tnikkel]

Approval Request Comment
[Feature/regressing bug #]: N/A
[User impact if declined]: May crash in low memory conditions.
[Describe test coverage new/current, TreeHerder]: The crash itself has only been observed in crash reports, not easily reproducible. However the crash occurs on a success path used every time an image is loaded and is thus well exercised by existing mochi/web-platorm/reftests.
[Risks and why]: Minimal, there was no functional change, only the need for some temporary allocations on the heap was removed.
[String/UUID change made/needed]: N/A

Comment 8

[Julien Cristau [:jcristau]]

Comment on attachment 8814389 [details] [diff] [review]
Remove string concat in ImageCacheKey::ComputeHash, v2 [carries r=tnikkel]

reduce risk of OOM, take in aurora52

Comment 9

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/35d579c4ff1e

Comment 10

[Gerry Chang [:gchang]]

Comment on attachment 8814389 [details] [diff] [review]
Remove string concat in ImageCacheKey::ComputeHash, v2 [carries r=tnikkel]

Fix a crash. Beta51+. Should be in 51 beta 5.

Comment 11

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/releases/mozilla-beta/rev/110c796d4372