New issue
Advanced search Search tips

Issue 787384 link

Starred by 1 user
Status: Fixed
Owner: ----
Closed: Nov 2017
Cc:
reillyg@chromium.org
hjd@google.com
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

Small regression in mini_installer size at 517578

Project Member Reported by hjd@google.com, Nov 21 2017 
CL [1] caused a small regression in size of on the mini_installer binary but looks like an improvement overall (see [2]) as mentioned in the CL.

Just opening as an FYI in case this wasn't expected.

[1]: https://chromium.googlesource.com/chromium/src/+log/89a01f157a882f2a8d7cc36d323ef6d62645ae29%5E..89a01f157a882f2a8d7cc36d323ef6d62645ae29?pretty=fuller&n=1000

[2]: https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgr--A6woM
Project Member

Comment 1 by 42576172...@developer.gserviceaccount.com, Nov 21 2017 

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=787384

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=e238daaf3406b70a5482d404b5df0165561cf5d65d83a1979642ff29c0905136


Bot(s) for this bug's original alert(s):

win-32

Comment 2 by hjd@chromium.org, Nov 21 2017

Status: Fixed (was: Untriaged)

Comment 3 by hjd@chromium.org, Nov 21 2017

Description: Show this description

Comment 4 by reillyg@chromium.org, Nov 21 2017 

There was a similar small regression on the transit size of the Android APK vs. the raw APK and on-disk size because by compressing individual objects in resources.pak we make the file smaller but also less compressable. The problem is that we have to trade off three things:

1) Download (compressed) size of the full install.
2) Download size of patch updates.
3) On-disk size of the full install.

The CL mentioned above improves (3) in a way that doesn't hurt (2) because it runs gzip in a mode that forces it to process files in chunks which is relatively friendly to the patch algorithm. This however hurts (1) because you get the best compression ratio by compressing the entire file in a single pass and compressed data is not further compressable.

It would be interesting to get statistics on whether we send more bits keeping Chrome up to date or supporting new users installs (or reinstalls). My intuition is that (2) and (3) are the things to optimize for.
►

Sign in to add a comment