New issue
Advanced search Search tips

Issue 683393 link

Starred by 1 user

Issue metadata

Status: Duplicate
Merged: issue 629332
Owner: ----
Closed: Feb 2017
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 3
Type: Bug



Sign in to add a comment

See if any icudata can be stripped that was needed for encoding detection

Project Member Reported by agrieve@chromium.org, Jan 20 2017

Issue description

We changed from ICU->CED for encoding detection:
https://codereview.chromium.org/2081653007

ucsdet_* from ICU are now no longer used, however, our .apk size grew by almost 100kb from this change.

Either
a) CED is that much larger than ICU, or
b) There are now unused parts of ICU that are not being removed.

I've verified that nm does not report any symbols containing ucsdet within libchrome.so
Perhaps there's some more parts of icudtl.dat that can be removed now?

Pointers:
http://userguide.icu-project.org/conversion/detection
http://userguide.icu-project.org/icudata#TOC-Reducing-the-Size-of-ICU-s-Data:-Conversion-Tables
https://cs.chromium.org/chromium/src/third_party/icu/README.chromium

 
FYI  Issue 629332  had tracked the size increase in Windows binary due to CED. Didn't try shaving off more ICU files aggressively - some tweak for VC++ compiler did the trick instead. Would be great if the size regression on Android is addressed.
Increase in apk size is about 90KB https://bugs.chromium.org/p/chromium/issues/detail?id=629332#c5

Did some research... AFAICT data for charset detection are not present in icudtl.dat. They are mostly embedded in the source code, which would have already been removed at linking phase.
Mergedinto: 629332
Status: Duplicate (was: Available)
Thanks for investigating! Hadn't see  bug #629332 , so happy to dupe into there. 

Sign in to add a comment