2025-08-22 10:25
michaelni changed the topic of #ffmpeg-devel to: Welcome to the FFmpeg development channel | Questions about using FFmpeg or developing with libav* libs should be asked in #ffmpeg | This channel is publicly logged | FFmpeg 8.0 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
00:11
rodgort has joined #ffmpeg-devel
00:26
iive has quit [Quit: They came for me...]
01:12
System_Error has joined #ffmpeg-devel
01:17
fjlogger has quit [Ping timeout: 264 seconds]
01:19
klaxa has quit [Ping timeout: 264 seconds]
01:21
klaxa has joined #ffmpeg-devel
01:23
messier81 has joined #ffmpeg-devel
01:32
fjlogger has joined #ffmpeg-devel
02:17
beastd has quit [Ping timeout: 260 seconds]
03:07
messier81 has quit [Quit: Konversation terminated!]
03:21
jamrial has quit []
03:41
Martchus has joined #ffmpeg-devel
03:43
Martchus_ has quit [Ping timeout: 265 seconds]
03:44
messier81 has joined #ffmpeg-devel
03:59
messier81 is now known as Messier81
03:59
Messier81 has quit [Quit: Konversation terminated!]
04:17
_whitelogger has joined #ffmpeg-devel
04:54
fjlogger has quit [*.net *.split]
04:54
rodgort has quit [*.net *.split]
04:54
hbbs has quit [*.net *.split]
04:54
any1 has quit [*.net *.split]
04:55
rodgort has joined #ffmpeg-devel
04:55
hbbs has joined #ffmpeg-devel
04:55
mkver has quit [Ping timeout: 256 seconds]
04:57
rossy has quit [Ping timeout: 256 seconds]
04:59
rossy has joined #ffmpeg-devel
05:00
fjlogger has joined #ffmpeg-devel
05:01
any1 has joined #ffmpeg-devel
05:05
secondcreek has quit [Read error: Connection reset by peer]
05:15
wyatt8750 has quit [Ping timeout: 245 seconds]
05:46
mkver has joined #ffmpeg-devel
05:51
wyatt8740 has joined #ffmpeg-devel
05:56
GewoonLeon has joined #ffmpeg-devel
05:56
wyatt8750 has joined #ffmpeg-devel
05:56
wyatt8740 has quit [Ping timeout: 240 seconds]
06:06
mkver has quit [Ping timeout: 256 seconds]
06:19
arch1t3cht1 has joined #ffmpeg-devel
06:29
bsFFFFFF has joined #ffmpeg-devel
07:01
MisterMinister has quit [Ping timeout: 244 seconds]
07:35
ngaullier has joined #ffmpeg-devel
07:40
HarshK23 has quit [Quit: Connection closed for inactivity]
08:07
ShadowJK has quit [Ping timeout: 250 seconds]
08:18
ShadowJK has joined #ffmpeg-devel
08:21
GewoonLeon has quit [Quit: GewoonLeon]
08:21
GewoonLeon1 has joined #ffmpeg-devel
08:23
GewoonLeon1 is now known as GewoonLeon
08:28
GewoonLeon has quit [Ping timeout: 240 seconds]
09:24
GewoonLeon has joined #ffmpeg-devel
09:41
System_Error has quit [Ping timeout: 272 seconds]
09:53
System_Error has joined #ffmpeg-devel
09:54
GewoonLeon has quit [Read error: Connection reset by peer]
09:54
GewoonLeon1 has joined #ffmpeg-devel
09:58
GewoonLeon1 has quit [Ping timeout: 264 seconds]
10:17
beastd has joined #ffmpeg-devel
10:18
<
kierank >
yes it's slow as hell
10:19
<
kierank >
to move between first page and pages
10:19
<
kierank >
but obviously the usual suspects feel its ok to gaslight this project about forgejo and the infrastruture in general
10:20
<
beastd >
kierank: i just tried here and the number were similar but the page was displayed fast (way faster than >1s)
10:24
<
beastd >
yeah did try moving between pages and can confirm it is slow'ish. numbers around a second seem to be accurate.
10:36
System_Error has quit [Remote host closed the connection]
10:38
<
beastd >
kierank: "slow as hell" seems exaggerated to me. Browsing MRs list on videolan vlc is pretty similar regarding performance
10:39
<
kierank >
the whole forjego propaganda was "github is slow"
10:39
<
kierank >
and low and behold when there is actual PRs forgejo is slow
10:40
<
kierank >
and also forgejo is run on a toy server
10:40
<
kierank >
never forget: 23:46:28 <•kierank> remember kids, in this project infra is a gaslighting exercise
10:42
<
beastd >
kierank: that was not my words. "gitlab is slow" was not my argumentation.
10:42
<
beastd >
and as i understood some people mentioned it especially regarding diff browsing.
10:42
<
kierank >
not saying it was your words
10:42
<
beastd >
but you are right browsing the PR list is slow'ish
10:47
GewoonLeon has joined #ffmpeg-devel
11:51
GewoonLeon has quit [Ping timeout: 240 seconds]
11:57
bsFFFFFF has quit [Ping timeout: 248 seconds]
12:01
ccawley2011 has joined #ffmpeg-devel
12:11
GewoonLeon has joined #ffmpeg-devel
12:16
jamrial has joined #ffmpeg-devel
12:19
<
kasper93 >
Unironically it was quite a weird argument "gitlab is slow", so we should "go forgejo". Because for sure it will be fast and without any bugs, as any small oss project...
12:23
<
kasper93 >
we have one repository, with 498 PRs of which 105 open and it already started to choke. I worry in a year when those numbers will be higher, it might really start being unusable
12:23
<
kasper93 >
Nicolas was right
12:24
<
BtbN >
I don't think this scales with the number of open PRs
12:24
<
BtbN >
or PRs in general
12:25
<
BtbN >
There is most likely some missing index in the DB somewhere, so it pulling all the info of the users makes the DB do inefficient stuff
12:25
<
kasper93 >
it scales with something
12:25
<
kasper93 >
number of user will also grow over time
12:26
<
BtbN >
like, the forgejo repo itself has 9k PRs, and is not 10 times slower
12:27
<
BtbN >
So I'm not overly concerned about this getting slower and slower
12:29
<
BtbN >
the forgejo repo on Codeberg has near identical load-times even for its PR page
12:29
<
BtbN >
so I this is most definitely some inefficiency in how it does its SQL queries for that page
12:31
minimal has joined #ffmpeg-devel
12:31
<
kasper93 >
probably scales with number of visible pr
12:32
<
BtbN >
yeah, I think so too
12:32
<
kasper93 >
it probably loads additional info about them in naive way
12:32
<
BtbN >
It's a usual "beginner SQL trap" that you just join everything together and then send one humongous query
12:33
<
BtbN >
looks good and elegant, but the result is that the DB has to iterate over all permutations of all joined tables, and suddenly it has to iterate millions of rows for a simple query
12:33
<
BtbN >
No idea if it's that, but it kinda "smells" like it
12:33
<
kasper93 >
main worry is that it is easy to ddos server with such high load entry point
12:33
<
BtbN >
if you want to DDoS the server, you just have a handful of bots clame configure...
12:34
* kasper93
updates script
12:35
<
BtbN >
Though I
_think_ it caches that?
12:35
<
BtbN >
But there's plenty of files to blame...
13:05
bsFFFFFF has joined #ffmpeg-devel
13:12
<
llyyr >
far bigger repos on codeberg lag less, so i don't think it's a forgejo problem
13:15
System_Error has joined #ffmpeg-devel
13:42
mkver has joined #ffmpeg-devel
13:45
ngaullier has quit [Remote host closed the connection]
13:50
ngaullier has joined #ffmpeg-devel
14:29
<
BtbN >
I did have no time to investigate this deeper
14:31
<
tmatth >
ah well for anyone who might have access that might be a decent place to start
14:38
<
BtbN >
It's too incoherent for that imo, LLMs can do better than that.
14:40
minimal has quit [Quit: Leaving]
14:43
<
another| >
maybe not a top tier AI?
14:44
<
BtbN >
It's most likely invalid anyway, since it's a feature request to FFmpeg, not my builds.
14:45
<
another| >
yes, obviously
14:47
Sean_McG has joined #ffmpeg-devel
14:51
realies has quit [Quit: ~]
14:57
realies has joined #ffmpeg-devel
14:59
ccawley2011 has quit [Ping timeout: 255 seconds]
15:11
System_Error has quit [Ping timeout: 272 seconds]
15:23
MisterMinister has joined #ffmpeg-devel
15:29
ccawley2011 has joined #ffmpeg-devel
15:59
<
kierank >
mkver: epic patch
16:01
<
mkver >
kierank: That was ironic, I presume?
16:02
<
kierank >
I think it's great removing inline mmx
16:04
<
mkver >
kierank: There seem to be a lot of places where I am simply baffled that there is no SSE code already.
16:08
<
mkver >
E.g. the size 16 xy2 hpel dsp functions would be natural candidates for it. Or the size 16 qpeldsp.
16:18
<
Sean_McG >
100% in agreement with kierank, it was well past time to junk those
16:31
System_Error has joined #ffmpeg-devel
16:33
ngaullier has quit [Remote host closed the connection]
16:33
ngaullier has joined #ffmpeg-devel
16:35
GewoonLeon has quit [Ping timeout: 264 seconds]
16:46
Sean_McG has quit [Quit: leaving]
16:55
GewoonLeon has joined #ffmpeg-devel
16:59
HarshK23 has joined #ffmpeg-devel
17:01
fjlogger has quit [Remote host closed the connection]
17:01
fjlogger has joined #ffmpeg-devel
17:07
<
BtbN >
tmatth: setting it to 0.5s does not produce anything at least
17:09
ccawley2011 has quit [Ping timeout: 248 seconds]
17:19
ngaullier has quit [Remote host closed the connection]
17:21
<
tmatth >
BtbN: ah that's too bad
17:21
<
BtbN >
It probably just means it's multiple quries
17:30
lexano has quit [Ping timeout: 260 seconds]
17:31
ccawley2011 has joined #ffmpeg-devel
17:43
lexano has joined #ffmpeg-devel
17:48
GewoonLeon has quit [Ping timeout: 252 seconds]
17:53
lexano has quit [Ping timeout: 264 seconds]
17:55
psykose has quit [Ping timeout: 265 seconds]
17:58
psykose has joined #ffmpeg-devel
17:58
psykose has quit [Remote host closed the connection]
18:05
lexano has joined #ffmpeg-devel
18:07
<
BtbN >
tmatth: yeah, I have mariadb itself now logging queries longer than 0.1 seconds
18:07
<
BtbN >
it logs nothing
18:15
<
tmatth >
it begs the question if the delay is strictly server-side then
18:17
<
tmatth >
but like you said I guess it could still just be a ton of short queries piling up
18:22
<
BtbN >
It's obviously server side
18:25
minimal has joined #ffmpeg-devel
18:48
ccawley2011 has quit [Ping timeout: 240 seconds]
19:21
<
Lynne >
could it be some VPN-related thing?
19:21
<
Lynne >
some roundtrip the frontend does
19:32
<
kasper93 >
iirc BtbN said server goes 100% cpu during the query, so seems to be compute related
19:32
<
BtbN >
one mariadb process does
19:34
<
Lynne >
odd, normally vaccuuming takes the most time
19:35
<
BtbN >
It's doing
_something_ that makes the DB be busy, but does not trigger any slow queries
19:36
<
BtbN >
not sure what you mean by vacuuming
19:36
<
Lynne >
might be a postgres term
19:45
bsFFFFFF has quit [Quit: bsFFFFFF]
20:03
<
kasper93 >
clearly forgejo needs to be rewriten in asm
20:05
psykose has joined #ffmpeg-devel
20:07
psykose has quit [Remote host closed the connection]
20:08
psykose has joined #ffmpeg-devel
20:44
BradleyS has quit [Read error: Connection reset by peer]
20:44
Thiru has quit [Read error: Connection reset by peer]
20:45
Thiru has joined #ffmpeg-devel
20:46
BradleyS has joined #ffmpeg-devel
20:59
quietvoid has quit [Read error: Connection reset by peer]
20:59
quietvoid has joined #ffmpeg-devel
21:11
microlappy has joined #ffmpeg-devel
21:15
microlappy has quit [Remote host closed the connection]
21:49
Compnn has joined #ffmpeg-devel
21:49
Compn has quit [Read error: Connection reset by peer]
22:04
Kimapr_ has quit [Remote host closed the connection]
22:05
Kimapr has joined #ffmpeg-devel
22:25
lexano has quit [Ping timeout: 256 seconds]
22:26
Kimapr has quit [Remote host closed the connection]
22:26
Kimapr has joined #ffmpeg-devel
22:38
lexano has joined #ffmpeg-devel
22:41
<
kasper93 >
thanks for that fix, I must have missed during grepping
22:58
mkver has quit [Remote host closed the connection]
23:02
mkver has joined #ffmpeg-devel
23:40
HarshK23 has quit [Quit: Connection closed for inactivity]
23:41
minimal has quit [Quit: Leaving]