Issue metadata
Sign in to add a comment
|
0.1% regression in resource_sizes (MonochromePublic.apk) at 565336:566473 |
||||||||||||||||||||
Issue descriptionThis regression happened while the diagnose-bloat system was broken, and is somewhere in a range of ~1130 commits. I'm running a stepped sweep to narrow it down, and will formalize this bug when the culprit is isolated.
,
Jun 13 2018
Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8943804289018957152
,
Jun 14 2018
Here are the biggest contributing 50-commit sub-ranges: From: cf453ab08ab1df252e6454128356dbf33344e43f To: 32c61869c629d800bd3e91fe9f7656aec93b425c MonochromePublic.apk_Breakdown (+81,927 bytes) +2,982 bytes Native resources (no l10n) size -2,860 bytes Zip Overhead +81,928 bytes Native code size -1 bytes Package metadata size -122 bytes Java code size MonochromePublic.apk_Specifics +82,211 bytes normalized apk size +81,928 bytes main lib size +284 bytes main dex size MonochromePublic.apk_Dex +1 entries fields +4 entries methods +1 entries types +1 entries strings From: 74be4bc85264530ca32901400e3a50f9bf60dae3 To: 05191f367005cb174169d804625de5c39dfc89ce MonochromePublic.apk_Breakdown (-39,824 bytes) +81 bytes Native resources (no l10n) size -2,107 bytes Zip Overhead +1,884 bytes V8 Snapshots size -39,820 bytes Native code size -4 bytes Package metadata size +142 bytes Java code size MonochromePublic.apk_Specifics -39,820 bytes normalized apk size -39,820 bytes main lib size +4 bytes main dex size From: df7770d8d2d9855eb05b8f6dc26099610bd7b1be To: 30c4e5bc3613142bb4024e7273432a82137177d4 MonochromePublic.apk_Breakdown (+57,525 bytes) -2,486 bytes Zip Overhead +88 bytes Non-compiled Android resources size +1,912 bytes V8 Snapshots size -107 bytes Native resources (l10n) size +37,088 bytes Native code size +1 bytes Package metadata size +78 bytes Native resources stored (l10n) size +8,795 bytes Java code size +12,156 bytes Compiled Android resources size MonochromePublic.apk_Specifics +73,602 bytes normalized apk size +37,088 bytes main lib size +16,048 bytes main dex size MonochromePublic.apk_Dex +67 entries fields +147 entries methods +19 entries types +63 entries strings
,
Jun 14 2018
,
Jun 14 2018
(note that the second range is a decrease)
,
Jun 14 2018
I'm starting a sweep on the first 50 commit range, with step size 5.
,
Jun 19 2018
I 2593575 See detailed diff results here: out/binary-size-results/diffs/e41b7c535994de0d415b4a2a13d4de93674f1c4e..32c61869c629d800bd3e91fe9f7656aec93b425c/diff_results.txt Diff Summary +69668 bytes MonochromePublic.apk_Specifics normalized apk size for range: 92bfde3a33ab261c8f38858c0dca28eb90444d4b..64e23b8c876dca16adc1812afe825ae7f90836f2 +12642 bytes MonochromePublic.apk_Specifics normalized apk size for range: 64e23b8c876dca16adc1812afe825ae7f90836f2..079a988ab5daf52e9541dd3041f4cf477fa8aa7b +13 bytes MonochromePublic.apk_Specifics normalized apk size for range: 995548dc9cc5d04dd62d56d6365f56a2af2c0e8a..302db4754d416e45eda6755b9ecdb0787e9c8490 +6 bytes MonochromePublic.apk_Specifics normalized apk size for range: 079a988ab5daf52e9541dd3041f4cf477fa8aa7b..995548dc9cc5d04dd62d56d6365f56a2af2c0e8a +3 bytes MonochromePublic.apk_Specifics normalized apk size for range: 23ea4334967e4841d7fad9b4f557e5fecd813c4d..e41b7c535994de0d415b4a2a13d4de93674f1c4e +0 bytes MonochromePublic.apk_Specifics normalized apk size for range: 302db4754d416e45eda6755b9ecdb0787e9c8490..038b85fb53805bc4e8a738d866ecc9e102813338 +0 bytes MonochromePublic.apk_Specifics normalized apk size for range: 038b85fb53805bc4e8a738d866ecc9e102813338..23ea4334967e4841d7fad9b4f557e5fecd813c4d -2 bytes MonochromePublic.apk_Specifics normalized apk size for range: e41b7c535994de0d415b4a2a13d4de93674f1c4e..32c61869c629d800bd3e91fe9f7656aec93b425c -85 bytes MonochromePublic.apk_Specifics normalized apk size for range: af684eb199c6ee1de118c54d5ea95af5e4a215bd..92bfde3a33ab261c8f38858c0dca28eb90444d4b
,
Jun 19 2018
+73716 bytes from: commit f0739e861b6debbe47b69938999af5b3e51347b6 Author: afdo-chromium-autoroll <afdo-chromium-autoroll@skia-buildbots.google.com.iam.gserviceaccount.com> Date: Fri Jun 8 21:39:18 2018 +0000 Roll AFDO from 69.0.3451.0_rc-r1 to 69.0.3453.0_rc-r1
,
Jun 20 2018
Splitting up the second 50-change interval: I 1509 See detailed diff results here: out/binary-size-results/diffs/17573c429dad1deee0c668f3933efc21e6b61a3a..30c4e5bc3613142bb4024e7273432a82137177d4/diff_results.txt Diff Summary +22941 bytes MonochromePublic.apk_Specifics normalized apk size for range: a60f7661bc3b3b2a9d8656e6b38d634c209538f1..bf19da64902f00f096fbf36650161e04a8c38cae +20434 bytes MonochromePublic.apk_Specifics normalized apk size for range: 273cf09720958bdc83b75d7596ae5e0b1988310f..a60f7661bc3b3b2a9d8656e6b38d634c209538f1 +12443 bytes MonochromePublic.apk_Specifics normalized apk size for range: f73dfaffea91c083fe003a96728df96d9e535493..b29235f115c67e2e3f5de82c6b7cb55d413a2710 +12207 bytes MonochromePublic.apk_Specifics normalized apk size for range: b29235f115c67e2e3f5de82c6b7cb55d413a2710..e7a16b17e4d73077514144a064793bf3e832ecad
,
Jun 20 2018
+27308 bytes from:
commit ccad6929368005f86e4a5183990eea2ca3708383
Author: Joe Mason <joenotcharles@chromium.org>
Date: Mon Jun 11 20:04:07 2018 +0000
Run chrome_cleaner_unittests on all Windows bots.
chrome_cleaner_gtests and chromium_win_fyi_gtests were removed in
https://chromium-review.googlesource.com/1093159 because they didn't
run anywhere. Add chrome_cleaner_unittests back and put it on all
Windows bots.
,
Jun 20 2018
+47742 bytes from: commit 4be0c0fe4ac71d2d08d261a8e5e2f64d6ec94ed9 Author: afdo-chromium-autoroll <afdo-chromium-autoroll@skia-buildbots.google.com.iam.gserviceaccount.com> Date: Mon Jun 11 19:59:00 2018 +0000 Roll AFDO from 69.0.3453.0_rc-r1 to 69.0.3455.0_rc-r1 -27308 bytes from: commit a60f7661bc3b3b2a9d8656e6b38d634c209538f1 Author: Farah Charab <farahcharab@google.com> Date: Mon Jun 11 20:04:03 2018 +0000 Finch: Add configs for waterfall testing.
,
Jun 20 2018
Two of the commits in question here are odd: +27308 bytes, then -27308, while neither change smells like a binary-impacting modification. These need to be verified.
,
Jun 20 2018
Interesting finding: I cleaned my results, and re-ran the +27K commit. This time, it yielded a 0-byte delta. After the last of the 27K results was detected, I tried to manually sync, and hit a prodaccess-related failure. So I wonder if the sync operation silently failed during the diagnose-bloat run, and yielded a pair of false results. I'll file a separate bug against this.
,
Jun 21 2018
,
Jun 21 2018
Closing this off, as the regressions trace to AFDO rolls. |
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by 42576172...@developer.gserviceaccount.com
, Jun 13 2018