25 Sep 2020 |
_hc | the solution there is to pre-crunch and commit that to git | 11:42:29 |
kitsunyan | Resource shrinker is a different think | 11:42:41 |
kitsunyan | *thing | 11:42:48 |
_hc | it might use similar methods, e.g. randomly trying things and choosing the best result by measuring | 11:43:07 |
kitsunyan | It shrinks xml resources, not tries to optimize pngs. | 11:43:20 |
_hc | binary xml != XML | 11:43:44 |
kitsunyan | No, we discussed here multiple times, its process deterministic. The only problem is that it tried to change files in apk with shrinked ones. | 11:44:14 |
kitsunyan | Which sometimes produced different apks on different machines, but with 100% same files inside apk | 11:44:49 |
_hc | what was the diff then? | 11:45:14 |
kitsunyan | AGP tries to be "smart": if shrinked apk size is larger (which is possible), it restored the original apk | 11:46:35 |
kitsunyan | So on some machines it produced shrinked apks, on other machines it didn't change anything | 11:47:21 |
kitsunyan | But the problem here not resource shrinker itself, but the fact that the process of modifying apk files was unclear | 11:48:14 |
kitsunyan | It happens on AGP side completely, so it just uses the same java version as gradle | 11:49:06 |
kitsunyan | So the problem is in java, but still I got different results with the same java 8 version, but on different OSes (debian stretch (used on buildserver) and ubuntu 20.4) | 11:51:02 |
kitsunyan | This is why I say it's still not clear to me how this process depends on environment | 11:52:03 |
Bubu | The forum will be down for maintenance for a few minutes in about 30 min.
| 13:24:53 |
Bubu | In reply to @bubu:bubu1.eu The forum will be down for maintenance for a few minutes in about 30 min.
All Done! | 13:54:11 |
cdesai | that was exactly 30 minutes! | 13:56:45 |
Bubu | I started a bit early ;-) | 13:57:38 |
wb9688 | cdesai: Not exactly, if you count the seconds… | 14:06:16 |
cdesai | wb9688: well I only have my client show hh:mm - vertical monitor so limited room horizontally | 14:07:47 |
wb9688 | Ah, my client shows seconds as well, even on my vertical smartphone ;) | 14:08:27 |
wb9688 | (I really can't use my smartphone horizontally lol) | 14:08:51 |
wb9688 | It was :54 vs :11 fwiw | 14:09:09 |
wb9688 | So a little big shorter than 30 min | 14:09:22 |
cdesai | Bubu: simply unacceptable | 14:11:05 |
wb9688 | Lol | 14:11:45 |
wb9688 | Or should I say: roflshmsfoaidmt | 14:12:26 |
| * cdesai had to look that up | 14:13:33 |
wb9688 | cdesai: Some people are surprised it's an actual thing. I just found it randomly on Wikipedia at some time | 14:14:50 |