freddietilley has quit [Ping timeout: 480 seconds]
x10z has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Begasus has joined #haiku
<Begasus>
g'morning peeps
<johnny_b>
morning
HaikuUser has quit [Ping timeout: 480 seconds]
<nephele>
Good one at that
<nephele>
Begasus: I'm ready to give up on the love2d patch, wrote some platform integration code and because of some issue haikuporter delete my patch for the 4th or 5th time .-.
<Not-85dd>
[haikuports/haikuports] sikmir 3d1152c - GPXSee: bump to 10.4 (#6740)
<woffs>
any hints how to make the wireless network ipw2200 on Thinkpad X41 more stable? After transferring some data it stalls and even de-re-enabling doesn't help, the configuration doesn't get finished and the default route isn't set
eroux has joined #haiku
ablyss has quit [Quit: Leaving]
ablyss has joined #haiku
HaikuUser has joined #haiku
HaikuUser has quit []
gouchi has quit [Remote host closed the connection]
illwieckz_ has quit [Ping timeout: 480 seconds]
OrangeBomb has quit [Ping timeout: 480 seconds]
gouchi has joined #haiku
bitigchi has joined #haiku
ClaudioM has joined #haiku
mmu_man has joined #haiku
Slor has quit [Quit: Coyote finally caught me]
<waddlesplash>
woffs: route not being set is a stack bug unrelated to wifi. happens the second time on any connection
<waddlesplash>
ipw2200 driver isn't that stable on FreeBSD either unfortunately, iirc. not sure we can do a lot
<waddlesplash>
Begasus: do open a ticket for that. is it reproducible? and what hrev
Cian has quit [Ping timeout: 480 seconds]
MajorBiscuit has joined #haiku
MajorBiscuit has quit []
<woffs>
thx waddlesplash
countryboy has joined #haiku
MajorBiscuit has joined #haiku
OrangeBomb has joined #haiku
countryboy has left #haiku [#haiku]
MajorBiscuit has quit [Quit: WeeChat 3.4]
freddietilley1 has joined #haiku
freddietilley has quit [Read error: Connection reset by peer]
gouchi has quit [Remote host closed the connection]
Cian has quit [Ping timeout: 480 seconds]
HaikuUser has joined #haiku
HaikuUser has quit []
ClaudioM has quit [Quit: leaving]
_Dario_ has joined #haiku
Jupp_S has joined #haiku
bitigchi_2 has joined #haiku
bitigchi has quit [Read error: Connection reset by peer]
phaedros has joined #haiku
Cian has joined #haiku
countryboy has joined #haiku
gouchi has joined #haiku
countryboy has left #haiku [#haiku]
Begasus_32 has joined #haiku
Jupp_S has quit [Remote host closed the connection]
<Begasus>
waddlesplash, on hrev55916 (nightly), not reproducable ... (running 32bit in VBox)
<waddlesplash>
Begasus: OK. don't bother reporting then. Due to a probably related issue I made changes in this codepath in hrev55937
<waddlesplash>
so, if you upgrade past that, and it still occurs, then open a ticket
<waddlesplash>
but otherwise, most probably it's already fixed
<phaedros>
Hello folks. I'm running Haiku on a Thinkpad X240, work's Ok. I was an Amiga dude in the beginning, then I saw BeOS. I got the i386 versions, thoose was great. But as a Unix/Linux dude since early 90's, it's good to have another OS to lurk with. Now I got HW too that like it, before that I was only using virtual box.
<Begasus>
k, thanks for responding, so far after the reboot it worked out ok, and just did a full rebuild for it (with download) in the same state it was before
<waddlesplash>
yeah I think the issue is the same race condition I solved earlier
AlaskanEmily has joined #haiku
<waddlesplash>
phaedros: welcome :)
<phaedros>
Thanks :)
<extrowerk>
is there any eason to build llvmpipe with llvm9 instead of llvm12?
<Begasus>
didn't want to update earlier just to check up on some of the open PR's :)
<Begasus>
phaedros, name sounds familiar
Jupp_S has joined #haiku
<Begasus>
extrowerk, commented on some of you PR's also on haikuports, could you check?
<x512[m]>
extrowerk: I use LLVM 12 in my Mesa build, including RADV.
<extrowerk>
i already fixed some ,thanks.
<extrowerk>
x512[m]: thx
<extrowerk>
Begasus: will do the rest too, but my time is limited.
<Begasus>
oops, my bad extrowerk , should check email before commenting here ;)
pch has quit [Remote host closed the connection]
<Begasus>
same here extrowerk ... time ... :(
<Begasus>
waddlesplash, I'm still a bit puzzled with the change you requested to the PR for pip (regarding python310/python3.10)
<waddlesplash>
Begasus: Haiku packages can have a . in their names
<waddlesplash>
so, "python310" can be "python3.10", which is much more apparent that it's version 3.10 not 310
<Begasus>
I know, but all other python packages don't mention them
<waddlesplash>
they should.
<waddlesplash>
it's a bit late for 3.9 unfortunately, though I guess we could do a mass rename with REPLACES()?
<waddlesplash>
but for 3.10 we should do this before everything starts using the wrong naming
<Begasus>
ok, then it's up to those maintainers then
<waddlesplash>
possibly worth noting this in policy somewhere...
<Begasus>
indeed
<jmairboeck>
is there a reason why python3 is still 3.7 and not the latest version?
<waddlesplash>
probably "python3" package should be derecated and go away
<phaedros>
Begasus: It's from the book Zen and the art och motorcycle maintenance :)
<Begasus>
we still have "python" that's using python2.7 afaik
<jmairboeck>
Begasus: but the python3 package is python 3.7.12
<waddlesplash>
jmairboeck: yes, and that will likely remain so until it's removed at this point unfortunately
<waddlesplash>
python's unfortunate decision to make source-compatibility-breaking changes within major versions means that we have to have packages for each point release separately
pch has joined #haiku
<waddlesplash>
maybe we should rename python3 "python3.7" to avoid confusion
<Begasus>
I think atm we are moving away from python 3.6 already, so a bit soon to drop 3.7
<waddlesplash>
does pkgman already offer you a choice when you ask for "python3"?
<waddlesplash>
or at least cmd:python3
<jmairboeck>
ah, so there are breaking changes between python 3.x versions? I didn't know that.
<waddlesplash>
yes, there are
<waddlesplash>
that's why we started with all this craziness in the first place
<Begasus>
cmd:python3 reveas the 4 python versions with pkgman
<Begasus>
extrowerk, could you remove/delete the twolame-0.3.13.recipe in the process, looks good to me
<extrowerk>
a question: in some cases if i select text in vision, the whole sentence starts to get shifted/elongated back-and-forth as i select further characters or as i deselect some. Have you also experienced this phenoomen? What can be the reason for this? The line which gets shifted/elongated seemingly doesn't include any wide-chars and i don't use any own-installed fonts, only Noto Sans Display.
<extrowerk>
Begasus: i can.
<nephele>
It's because of subpixel alignment iirc
<extrowerk>
nephele: any way to eliminate it? interestingly not every sentence triggers this.
<nephele>
That is the width of the string "blah" is not the same as the width of b + l + a + h and such
<Begasus>
soname didn't change so it doesn't effect other dependencies (afaikt)
<nephele>
Use gnu unifont /s, but yes it is known, i am not sure how to really fix it
HaikuUser has quit []
<extrowerk>
Begasus: done
<nephele>
Why are you using Noto Sans Display for Vision though?
<extrowerk>
nephele: is that a font?
<nephele>
Yes it is, but i don't like it aestethically
<extrowerk>
"GNU Unifont is a free Unicode bitmap font" No thanks.
<nephele>
I would try using Noto Sans instead of Noto Sans Disply to work arounf this
<nephele>
I don't understand why you use the Display variant in the first place
<extrowerk>
nephele: because Noto Sans looks terribly bad
<nephele>
... why? It's the same as Display just that display tries to be less wide
<Begasus>
seems like a conflict now extrowerk
<extrowerk>
nephele: yep, the Sans variant is too long, have too much space between the chars.
<extrowerk>
eg: looks bad
<Begasus>
maybe your branch wasn't on par with master?
<nephele>
It's ment for menus
<extrowerk>
Begasus: possible.
<nephele>
anyhow, i guess, upvote the ticket, or work on app_server ;)
<extrowerk>
nephele: it seems Comic Sans doesn't triggers this bug.
<nephele>
There are lots of fonts that don't trigger this bug
<extrowerk>
Oh, this MT extra lloks also terrible.
<Begasus>
now check up on buildmaster/reports
HaikuUser has joined #haiku
HaikuUser2 has joined #haiku
<HaikuUser>
hello
<Begasus>
hello
HaikuUser has quit []
HaikuUser2 has quit []
<Begasus>
extrowerk, on gettext, as mentioned, maybe try to keep the previous version in some kind and update to the newer version?
<Begasus>
would be nice to get this out of the way also ;)
<jezek2>
extrowerk: the shift is caused by kerning, and the easiest way to eliminate it is how the text is rendered, eg. using clipping to render the highlighted part, but you can also directly work with the kerning
<nephele>
Uhh, what do you mean by caused by kerning?
<jezek2>
also thinking about it, many APIs offer just integer coordinates so that's also the cause
<jezek2>
as the characters have fractional widths
<jezek2>
kerning is a different spacing between some pair of letters
<nephele>
We support rendering this fine, but that isn't the reason for the misrendering when selecting
<nephele>
that is just an API problem basically
<jezek2>
I would bet that the API is using integer coordinates and/or the kerning is not properly accounted for when calculating text width
<jezek2>
but even without such things, with clipping it's possible to easily fix it, just render the same text (at the same position) twice, one time the non-selected part, then the selected part by clipping
<nephele>
No, the api is not using integer coordinates
<nephele>
What is the point of guessing? we already know the cause :/
<jezek2>
which is?
<nephele>
Uhh, did something change in trac? I can't add filters to my querries anymore g:
freddietilley1 has quit [Ping timeout: 480 seconds]
<Begasus>
all good extrowerk (on twolame) :)
Cian has quit [Ping timeout: 480 seconds]
<nephele>
jezek2: I wrote above "That is the width of the string "blah" is not the same as the width of b + l + a + h and such"
<nephele>
iirc that is the problem, that callers expect this to be true
<nephele>
that is you can split the string arbitrarily and render parts of it after the other by using the returned positions
Begasus_32 has quit [Quit: Vision[]: i've been blurred!]
<nephele>
I can't find the ticket currently... trac somehow is broken, I think PulkoMandy explained it better on the ticket :/
<DHowett>
but that's because of kerning. You said "that isn't the reason for the misrendering when selecting", but if you're saying that the reason for misrendering when selecting is that "the width of the string \"blah\" . . . " then it IS because of kerning?
<DHowett>
o_O
<nephele>
No, it isn't caused by kerning
<nephele>
it's also stuff like ligatures, and subpixel coordinates, it's really ab api problem
<Begasus>
done for today, g'night peeps
Begasus has quit [Quit: Ik ga weg]
<jezek2>
subpixel positioning should be fine if everything is in floats (and no rounding is occurring somewhere sneakily)
<nephele>
kerning is the width between characters, it is a property of fonts, but that this exists isn't the reason for this bug, it would still exist if we had monospace fonts with ligatures :)
<nephele>
also, this assumption completely breaks if you have something like arabic, but we don't support this anyway currently
<jezek2>
I hate ligatures, maybe that's why I don't never think about them :P
<DHowett>
monospace fonts with ligatures are ill-formed if the ligatures violate the 1:1 width constraint from codepoints to columns
<x512[m]>
Some font rendering systems use subpixel coordinates, but align glyph by pixels when drawing.
<DHowett>
if the API is incapable of handling such a font, then yeah, it's the API's fault
<nephele>
1:1 width?
<nephele>
monospace fonts don't gurantee the same size glyphs either, you already have to deal with full widrh or half width glyphs
<nephele>
x512: Yeah we definetely respect font hinting, but this shouldn't pose a problem
<DHowett>
a defining characteristic of a monospace font is that each glyph's bounding box is the same or an appropriate integer multiple depending on how many glyphs are composed into it
<nephele>
DHowett: it's probably more that the caller is using the api wrong, but anyhow this goes into specifics that are hard to reason about without the ticket at hand
<DHowett>
indeed
<DHowett>
(sorry: I'm passionate about these things as the engineering owner of both a monospaced font and a terminal emulator that has to deal with this crap ;P)
<x512[m]>
Monospace do not directly related to pixels.
<nephele>
Yes, but ij would be two boxes big but ij as a ligature would be one box big
<nephele>
so if you render it at the boundary you would ger two boxes, and all at once would be one box
<x512[m]>
It is just the same width of most characters.
<DHowett>
right but if a monospace font ligates i j into a single-cell-width ij, it is not a monospace font :)
<DHowett>
it is a duospace font at best
<nephele>
there is a ij standalone unicode glyph also ;)
<DHowett>
which is fine, and should be a single cell width
<x512[m]>
CJK takes at least 2 monospace character width.
<nephele>
That's called full width, yeah
<nephele>
ij vs ij :), confusing users trying to use their text cursor hehe
vezhlys has joined #haiku
<nephele>
If you want to use when to use this glyph in a correct way gramatically iirc you would need to parse sylables of dutch, and not use it if it is ob a sylable boundary (which might happe because dutch doesnt change the spelling of imported words)
<nephele>
uhh, i should stop my linquistical tangent :)
freddietilley1 has joined #haiku
dcatt has joined #haiku
<dcatt>
Hello fellow Haikuians! :D
<nephele>
Hello dcatt
freddietilley1 has quit [Ping timeout: 480 seconds]
AlienSoldier has joined #haiku
<augiedoggie>
kallisti5: i have the same crash as #17377 but with a different card and possibly a different reason for the crash
eroux has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Cian has joined #haiku
kevans91 has quit [Quit: Reconnecting]
kevans91 has joined #haiku
kevans91 has quit []
kevans91 has joined #haiku
illwieckz has joined #haiku
lorglas has joined #haiku
dcatt has quit [Quit: Vision[]: i've been blurred!]
<nielx[m]>
extrowerk: what's the error? I have had reasonable success cross-compiling llvm as part of rust
<extrowerk>
just building it as i deleted my work directory for some kind of reason.
<extrowerk>
i however already created a PR with my wip recipe and patchset, as soon as the build fails again, i will quote the error message in the PR comment.
<extrowerk>
hope it will be useful for somebody.
lorglas has quit [Quit: Vision[]: i've been blurred!]
<nielx[m]>
please @ me
<extrowerk>
done
bitigchi_2 has quit [Ping timeout: 480 seconds]
<nielx[m]>
thanks, I'll keep my eye out
HaikuUser has joined #haiku
HaikuUser has quit []
bitigchi_2 has joined #haiku
jmairboeck has quit [Quit: Konversation terminated!]
siguri has quit [Quit: Vision[0.10.3]: i've been blurred!]
gouchi has quit [Remote host closed the connection]
bitigchi_2 has quit [Ping timeout: 480 seconds]
<extrowerk>
does the LLVM port really have to support _everything_?
<x512[m]>
extrowerk: ?
_Dario_ has quit [Quit: Vision[]: i've been blurred!]
<extrowerk>
x512[m]: it seems the llvm recipe enables all the possible targets. This is not necessarily required and disabling some can radically reduce the compilation time / ram usage
<x512[m]>
This may help to compile LLVM for RISC-V. Only riscv64 and Radeon GPU targets are needed.
<extrowerk>
I am looking at this blogpost-series (check archive, if interested): https://www.cambus.net/speedbuilding-llvm-clang-in-5-minutes/ I understand, to achaive this one needs tons of CPU cores and huge amount of ram, but we can still consider to disable some parts of llvm to make the compilation quicker.
<extrowerk>
x512[m], on your screenshots Haiku is a mix of english and chineese or japaneese (i am unable to read those moonrunes, so for me all of them looks the same, sorry). Are you a multi-language person?
<x512[m]>
I born and spent childhood in Russia and currently living in Japan. I know 3 languages: Russian, Japanese and English.
AlaskanEmily has quit [Read error: No route to host]
AlaskanEmily has joined #haiku
bitigchi_2 has joined #haiku
bitigchi_2 has quit [Ping timeout: 480 seconds]
<extrowerk>
x512[m]: please repeat what arch support you need for radeon in llvm.
<x512[m]>
I have not checked exactly, but it seems that LLVM is used to compile to shader bytecode.
<x512[m]>
It just works so I have not checked details.
<extrowerk>
I assume AArch64,, AMDGPU, ARM,PowerPC,RISCV,Sparc,WebAssembly,X86, can be enough for us. MAybe even without PowerPC and Sparc
<extrowerk>
reduced the list to "AMDGPU;RISCV;WebAssembly;X86"
Therion has quit [Quit: Vision[]: i've been blurred!]
<extrowerk>
Now make have 1920 targets to build instead of 5000+.
<x512[m]>
The best would be add-on architecture with dynamically loadable architecture add-ons.
bitigchi_2 has joined #haiku
<x512[m]>
So it can be built and packaged separately.
<extrowerk>
you mean to split the llvm package to specific arch subpackages?
<extrowerk>
*arch specific*
<x512[m]>
Yes. But not like GCC where each architecture should be compiled fully independently and not interact al all with each other.
<extrowerk>
looking at the provided libs it seems each arch got their own libs. it seems to be possible to split the package to subpackages. but it takes ages to build llvm on my computer, so i won't really do that.
<augiedoggie>
i would prefer if you didn't remove all of the microcontroller targets