| 14 Sep 2021 |
 _hc | FYI: https://gitlab.com/fdroid/fdroiddata/-/merge_requests/9770 | 14:03:46 |
 jochensp | _hc: not sure if that is the best option, did you test it? | 14:04:57 |
 Izzy | jochensp: like that? Unfortunately I saw your comment only after having sent the "official toot". So I amended my own with details for your incredible work on that. | 14:05:13 |
| _hc | yes, I tested it, and its running a build in CI | 14:05:46 |
| _hc | built super fast on a VM with 32GB of RAM :) | 14:06:08 |
| jochensp | :D | 14:06:22 |
| Izzy | First ovations received :D | 14:06:46 |
| _hc | on CI, it only get 2G allocated | 14:07:04 |
| _hc | why on earth doesn't Java/Gradle just let you say: "use all avilable RAM". It feels like i'm stuck in the 90s to have to manually set limits like that | 14:17:11 |
 tepozoa | systems tech here, java will definitely use all RAM in sight if you let it just run at defaults. sounds like a specific recipe issue | 14:18:30 |
| _hc | maybe its gradle's fault | 14:19:50 |
| jochensp | as far as I read it is about the heap memory: https://stackoverflow.com/questions/14763079/what-are-the-xms-and-xmx-parameters-when-starting-jvm | 14:20:45 |
| _hc | why would I then have to mess with these settings: "org.gradle.jvmargs=-Xms1g -Xmx4g -XX:MaxPermSize=4g" | 14:20:52 |
| jochensp | (and I still don't think allocating all memory to the heap is correct) | 14:21:26 |
| tepozoa | yeah, the heap is not where the classes, methods and other objects are stored - that's the MaxPermSize | 14:22:10 |
| tepozoa | your heap is basically a lot of pointers and whatnot (I'm not an actual programmer, go light on me here) - for example, when we moved from 32bit JVM to 64bit JVM it require(s)(d) increasing the heap allocation by about 30% to account for the increased width of data structures - but 1.5G is usually plenty | 14:23:50 |
| tepozoa | https://www.oracle.com/java/technologies/javase/vmoptions-jsp.html | 14:24:50 |
| tepozoa | I've learned over the long term is super easy to mis-tweak the JVM with well-intentioned hand tuning. The JVM is scary good at usually dynamically figuring a lot of it out on it's own | 14:26:03 |
 bene64 | Wow, that Element patch really landed fast. Congratz to everyone involved β€οΈ | 14:38:07 |
 εΉΈη« (πππΎπ/πππΎπ) | fyi: "Element would've been willing to share our private source tree for Element Android 1.2.2 with F-Droid several days in advance of disclosure, if it meant that they could pre-build the binaries and have them ready to ship the moment that the source was made public." https://news.ycombinator.com/item?id=28514809 | 15:36:41 |
| εΉΈη« (πππΎπ/πππΎπ) | In reply to @_oftc_jochensp:matrix.org
probably echo "\n.." would work I tend to always use that since there are a lot of non-posix-compliant text files out there these days because several "modern editors" treat newline as a separator instead of a terminator; POSIX mandates it as a terminator, so things break. | 15:39:05 |
| jochensp | yeah.. | 15:39:24 |
| εΉΈη« (πππΎπ/πππΎπ) | I noticed several updates that were built before the element security fix are "missing"; e.g. the german covpass apps (which were in a batch started sunday morning I think); the dutch apps are now published though, even though they were in the next batch. | 15:42:41 |
![@_oftc_Licaon_Kter[xmpp]:matrix.org](./avatar/Licaon_Kter%5Bxmpp%5D) Licaon_Kter[xmpp] | εΉΈη« (πππΎπ/πππΎπ): it was a "stop the press" moment | 15:44:33 |
| jochensp | I think the batch from Sunday is somehow lost, unrelated to the "stop the press" | 15:44:54 |
 cde | I guess this may be easier when there's multiple buildserver support - but it'd be good to have a way to just do a quick one-off build. | 15:45:03 |
| jochensp | there was a new build run yesterday and that seems to be included | 15:45:18 |
| cde | and I'd argue it'd be good to use that for fdroidclient itself as well | 15:45:21 |
| cde | In reply to @cdesai:matrix.org
I guess this may be easier when there's multiple buildserver support - but it'd be good to have a way to just do a quick one-off build. more of a "pause things" rather than 'stop", or just parallel building, which is easier and works. | 15:45:55 |
| jochensp | ..and it is not the first time we see build runs getting lost and only the rebuild getting published later, but I have no idea how that could happen | 15:46:21 |
