PhorgePublicGeneral discussions about Phorge. For long and complex questions, you may want to use Ponder.
Thanks avivey! And enjoy in production the new filter by token... lol @taavi https://we.phorge.it/token/given/query/imgpytwdtshH/#R
Given we have no release this year yet, this is an opportunity to decide on which version naming we should use: YYYY.ww (YYYY Week ww) or YYYY.mm (mm/YYYY). Until 2015, Phabricator was using YYYY.mm before switching to weekly releases. However, we (Phorge) promote to stable only several times per year.
@avivey: maybe phd is not running
@aklapper Honestly, the current naming scheme looks fine to me, it's just that Phorge stable releases are several times a year, not weekly (Phab was weekly from 2015 to 2021 with some exceptions) and Phorge is unlikely to have two releases per month (rare exception: security vulnerabilities).
During the beginning (when 2022.35 got promoted), I though Phorge may have weekly releases one day and today I think some may be thinking that Phorge releases could be weekly.
@Cigaryno: Ubuntu releases, twice a year, end in .04 and .10. They are not released monthly. So I guess I don't mind if people think that Phorge releases could be weekly. If they want to find out they could go to https://we.phorge.it/w/changelog/ :)
In Z1#3558, @aklapper wrote:
@Cigaryno: Ubuntu releases, twice a year, end in .04 and .10. They are not released monthly. So I guess I don't mind if people think that Phorge releases could be weekly. If they want to find out they could go to https://we.phorge.it/w/changelog/ :)
Agree with you. Staying on YYYY.ww despite little releases is therefore the equivalent of Ubuntu release twice a year and using .04 and .10 (for April and October).
The new user @Chlod is a trusted user, thanks admin for accepting ๐
Looks like https://secure.phabricator.com is dead? Receiving 503. Which is a shame because the old tasks/revisions is very useful for documentary purposes.
When I last checked, the amazon loadbalancer reported 503 Service Unavailable: Back-end server is at capacity; it has been having issues with full diskspace for a while. It's probably AI scraper bots that cause a flood of temporary files from commits and revisions. With some luck the garbage collector gets to them in a few days.
Aaaand here a Discussion Needed thing maybe interesting, maybe already discussed in secure dot (?): T16060: arc diff: it could auto-claim the Task, if unclaimed
You should have a "Create security task" link from the favorites (flag next to your profile picture on the top) that takes you to https://we.phorge.it/maniphest/task/edit/form/1/
It's possible that this form isn't configured properly though.
Could a Blessed Committers take a look at D25955? It would make development a lot easier if that is landed.
I upgraded from Fedora 41 with PHP 8.3 to Fedora 42 with PHP 8.4. Now both Firefox and Chromium only display a "Content Encoding Error" instead of local Phorge. After applying D25887: Remove use of deprecated E_STRICT PHP constant Phorge works again.
Urgh... I cannot even arc diff anymore on PHP 8.4: https://we.phorge.it/D25887#26909
Uhm maybe we need a ./phorge/bin/diviner generate, from arcanist
Whooo thanks avivey โญ woooo, finally example usages about X.$ and JX.DOM.scry and JX.DOM.find and JX.DOM.findAbove with jQuery-like examples wooo look mum I know Javelin! :O https://we.phorge.it/book/javelin/article/javelin_intro/
Could Blessed Committers take a look at D25872? @valerio.bozzolan
It turns out that there's a broken importer in Nuance - the GitHub importer - that is broken, and that floods the Daemonsโ log, which we keep in DB.
That log was expanding faster then the log GC was able to clean it up, and got to around 66 million lines - about 14GB.
Someone also tried to upload a 10gb file a few weeks ago, but I'm not sure how much actual bytes that took. I removed it before checking.
anyway, those were enough to consume our entire disk space, after which Phorge can't connect to the DB and crashes.
And since the bad connector was still up, we were generating garbage logs.
Now, I've stopped Nuance on this install, and I suggest you do the same on other installs. I'll file a ticket and probably just kill the bad code, assuming it doesn't work anyway.
I've also manually cleared the daemon_logevent table, which was over inflated, and the DBโs binlogs, which is another thing that takes lots of space.
Also took the opportunity to look for any possibility-evil files in the DB and only found the large one mentioned above.
And there may be a problem in the log GC code, I'll check more closely.
I'm kind of conflicted about whether it makes sense to keep logs in the db in 2025. I'm a fan of systemd journal and https://grafana.com/oss/loki/
Avivey - Just out of curiosity, are what are the errors from the GitHub importer? - Something like :"Failed to open stream: HTTP request failed! HTTP/1.1 429 Too Many Requests" ??
They look like this:
<cursor:events.repository> Polling GitHub Repository API endpoint "/repos///events".\n [2025-02-10 22:16:47] EXCEPTION: (Exception) You must setAccessToken()! at [<arcanist>/src/future/github/PhutilGitHubFuture.php:49]
with different urls.
it looks like it didn't even make the rest call...
Uhm there was a document in Diviner about "please first commit abstract things, then commit their usages". Maybe in the document 1 idea = 1 thing. I cannot find it
Uhm. Should a page like /file/edit/1/ be HTTP 401 from non-logged in? Instead of showing an HTTP 200 login page?
I ask just because I was lazily trying to block some very annoying AI bots but I cannot easily distinguish the ones hammering the login page as logged-out, from the ones that are just doing their job as logged-in, from the traditional Apache log
I guess it depends by the need
I know at least at my work, having See Txxx outright attach commits rather than merely mark them as mentions would sometimes lead to unintentional results, so I'd selfishly argue a bit against that. Though, either way I would reckon that the additional special syntax "documented" in https://web.archive.org/web/20250321014228/https://secure.phabricator.com/T5132 should finally make its way into https://we.phorge.it/book/phorge/article/remarkup/.
P.S. about documenting "Ref Txxx" stuff, here task https://we.phorge.it/T15153
Interestingly T16085 is reported as "This task is connected to more than 200 other tasks" ... but not really lol https://we.phorge.it/maniphest/graph/16085/
@valerio.bozzolan at a guess, maybe it's referring to everything that's attached to T16085's parent task - https://we.phorge.it/maniphest/graph/15064/
Gift idea: a small 3d model "Phorge Stable is Out ๐ป"
Hey, I temporarily removed myself from Trusted Contributors in order to test something - please could somebody add me back? :p