<nephele>
should I just disable swap if my build is hitting my memory limit?
<nephele>
(so it does not swap thrash) or would that not be a problem normally?
freakazoid332 has quit [Ping timeout: 480 seconds]
B2IA has quit [Quit: Vision[]: i've been blurred!]
B2IA has joined #haiku
<waddlesplash>
nephele: I have swap disabled since I'm on a ssd
<nephele>
This is also a ssd, and not a big one either... Though my webkit compile pretty quickly scrapes the llimit of my memory
<waddlesplash>
use Clang
<waddlesplash>
and disable ASLR
<waddlesplash>
like the HaikuPorts recipe does
<waddlesplash>
saves a bunch of memory
<nephele>
the haikuports recipe uses clang already?
<nephele>
I'll check out how it compiles it
<nephele>
waddlesplash: gcc is in build_requires, is that correct?
<waddlesplash>
I think the gcc package is also needed?
<waddlesplash>
can't recall
<waddlesplash>
but the latest recipe definitely uses clang
<nephele>
no thumbnail view for IOM files, sad
<nephele>
waddlesplash: neat. thanks
<nephele>
i assume that pkgconfig line is not required though
<nephele>
lol, we have a firefox icon
nephele has quit [Ping timeout: 480 seconds]
deneel has joined #haiku
<Skipp_OSX>
long requested...
nephele has joined #haiku
<nephele>
Skipp_OSX: in the icon directory i mean
<nephele>
I just spend like half an hour starting a rEFInd theme, rebooted the system to check it out... and forgot to remove the "boot in text mode" option in my config :D
<nephele>
also, now my rEFInd has a scaled 64x64 png haiku cursor... but my actual haiku cursor is stuck at non-scaled, so that is funny :D
<Skipp_OSX>
hrev28136 apparently steppe added it
<Skipp_OSX>
stippi
<nephele>
might have been for the past firefox port
AlaskanEmily has joined #haiku
deneel1 has joined #haiku
<nephele>
clang seems to be more picky than clang, already failing the compile :(
<nephele>
than gcc*
deneel has quit [Ping timeout: 480 seconds]
<nephele>
it's erroring out in included headers from haikus gcc 13 .-.
<nephele>
c++/bitset
<nephele>
seems to be "only" in jshell, and so far i can skip that in the build
<nekobot>
[haiku/haiku] 10bba152b3d5 - bootloader: Set a default heap size for all platforms.
<nekobot>
[haiku/haiku] 38946ff82f4d - bootloader: Increase the "large allocation" threshhold to 128 KB.
<waddlesplash>
that second commit fixes a really long-standing memory leak in the bootloader that's bled over into the kernel
<waddlesplash>
on my system it reduces post-boot memory usage by over 100 MB
frkazoid333 has joined #haiku
<nephele>
is that dependant on packages active? that seems quite significant
<waddlesplash>
nope
<waddlesplash>
or at least I don't think so, I think it depends only on the Haiku package itself
<nephele>
well, the last test for haiku booting was in the 256mb range... 100mb reduction to that is quite a bit
<waddlesplash>
yes
frkzoid has joined #haiku
frkzoid has quit []
HaikuUser has joined #haiku
frkazoid333 has quit [Ping timeout: 480 seconds]
<nephele>
WTFString has no == operator with const* char *; nice. Atleast clang thinks so "invalid operands to binary expression ('String' and 'const char *const')
rjwiii has quit [Ping timeout: 480 seconds]
<nephele>
which begs the question... why did gcc think this was okay? *shrug*
HaikuUser has quit []
<waddlesplash>
nephele: there are patches in HaikuPorts for builiding HWK with Clang
<waddlesplash>
which I don't think PulkoMandy merged to the HaikuWebKit repo yet
<nephele>
lol, i'm confused why you fixed that same error with a StringView, while below a String from wk is returned... I picked a different fix in my case. I guess it does not matter too much
<nephele>
ah StringView is also a webkit type
<waddlesplash>
yes
<waddlesplash>
this is what should be used for const data
<waddlesplash>
it avoids copying the const data
<nephele>
Does the fromLatin1 instead of fromUTF8 also pertain to that?
<nephele>
for the extension map I'm not sure if you can be sure if there is something that isn't within latin1
<waddlesplash>
that's fair
<waddlesplash>
is there a fromUTF8? I remember just copying what other code did
<nephele>
Yes there is, atleast for String. maybe not for StringView?
<nephele>
hmm, the other two fixes for Webkitlegacy and dumrendertree are not relevant to me right now since i'm building neither
<nephele>
and your patch corrects some whitespace :)
<waddlesplash>
might as well merge the whole patch
<nephele>
it might make sense to adjust the haikuports recipe to not bother compiling DumpRenderTree and Jsshell however
<nephele>
don't merge whitespace changes though, that gives uneccesary difference to upstream. Apart from the latin1/utf8 for mime your patchset looks good anyhow
<nephele>
wondering why you had to tell egl not to use x11... do you have headers on your device it picked up on? with gcc i didn't have any problems with that so far
<nephele>
memory usage with clang and aslr disabled is much lower now, about 20GB with 12 threads compiling
<waddlesplash>
nice
<waddlesplash>
I can't recall exactly why I needed those hacks. I remember tracing it back to something in Mesa that was actually different for Clang but maybe I fixed that?
<nephele>
Well, I'll see if this crops up
<nephele>
In wk2 *afaik* egl should not be currently needed? but i'm not too sure about that
<nephele>
heres hoping that I'll have a working debug build in some hours...
nephele has quit [Quit: Vision[]: i've been blurred!]
OscarL has joined #haiku
<OscarL>
re: hrev58211 / #14831... nice work waddlesplash! :-)
<waddlesplash>
:)
tuaris has quit [Remote host closed the connection]
<nipos>
Not sure if I reported this one before. Happens on shutdown on my Asus Transformer Book T100TA. Not that important, as I can't use it due to missing network and internal storage anyways, but maybe some day :)
tuaris has quit [Quit: Leaving.]
<yann64>
Hi, is theer a way to declare a persistant environnement
<yann64>
variable in a receipe?
<Begasus[m]>
as in?
<Begasus[m]>
should be some examples I guess
<yann64>
the binary I am packaging need an environement variable to be set permanently
<yann64>
looking in the receipes but so far could not fin an example
<Begasus[m]>
it needs the variable after packaging/installing?
<yann64>
tried adding "export MYVARIABLE=test" in teh install section of the package, but once the package installed, doint "echo $MYVARIABLE" does not output "test"
<yann64>
yes, persistant once package is installed
<Begasus[m]>
don't think that's possible, but could be wrong, never seen it happen
<Begasus[m]>
one way (maybe) would be to use a post-install script that adds a line in ~/config/settings/profile? but don't think it's good practice as the user/installer can't control that
<erysdren>
i think some gcc recipes do something like that?
<erysdren>
not sure
<yann64>
good point, will check if something like that is done with gcc
<erysdren>
once upon a time i tried to port DJGPP, though i can't remember if i figured out a way to persistently set DJDIR...
<erysdren>
shrug.
ChaiTRex has quit [Remote host closed the connection]
ChaiTRex has joined #haiku
<phschafft>
other packages on other systems normally use a shell script/batch job to set up things before the actual binary is started.
<phschafft>
if you do so, keep in mind to use exec to start the binary.
janking has joined #haiku
<yann64>
the script approach looks doable albeit compilcated, will investigate this
<yann64>
or I could tell the install script to add a "export MYVARIABLE=test" line to /boot/system/settings/bashrc , but that is not very clean and portable to a multi-user
<phschafft>
yes. and think about it: if every package would do that....
<phschafft>
also how to maintain that in case the value changes at some point?
<phschafft>
even just detecting if it is set to the correct value sounds like a nightmare.
<Monni>
sed can grab the current value... it can also change it...
<phschafft>
yes, and if the correct value is something like 'true'....
<Monni>
When comparing the values have to be quoted...
<phschafft>
sed-ing other people's configs is like doing cardiac surgery with a glowing hot sword and your eyes closed.
<Monni>
sed is used in a lot of CMake scripts...
<phschafft>
so what?
<phschafft>
I do not suggest that sad is a bad tool. I suggest that what you suggest is a very fatal idea.
<Monni>
If done correctly, there is very small chance that it will change something that it isn't supposed to...
<phschafft>
'If done correctly, ' has always evaluated to false in reality.
<Monni>
Not true
<phschafft>
someone who claims to be like 40 years of doing this should really know that.
<phschafft>
please just stop giving bad, or even wrong advices
<Monni>
That's your opinion, not fact...
<gordonjcp>
yann64: what's the package/binary? What's the variable for?
<phschafft>
I think it is sufficiently established that if you implement something non-fool-prove it will fail eventually.
<phschafft>
Also even with it in the rest says the same. editing user config in package setup is just the wrong thing to do.
<gordonjcp>
if you do implement something fool-proof someone will find a way to break it all the same
<phschafft>
plus there is the established way of doing it via a proxy script.
<phschafft>
gordonjcp: yes, but you can make it so that it will break after you found a new maintainer for the package! ;)
<gordonjcp>
phschafft: that goes
<gordonjcp>
phschafft: I work for the Fire Brigade
<gordonjcp>
phschafft: nothing is firefighter-proof
* phschafft
nods.
<gordonjcp>
you might make it firefighter-resistant, which means it could last for months before they properly break it
<gordonjcp>
we use the same kind of buttons they use on industrial factory control gear for things like "Reprint last turnout" or "Local Turnout" buttons
<gordonjcp>
they're a hefty chunk of stainless, with a polycarbonate (motorbike helmet plastic) button, and a glass-filled nylon switch body, with a V3-type microswitch inside
<gordonjcp>
you can hit them with a hammer as long as you like
erysdren has quit [Quit: Konversation terminated!]
yann64 has quit [Quit: Vision[]: i've been blurred!]
Begasus has joined #haiku
vdamewood_ has joined #haiku
vdamewood has quit [Read error: Connection reset by peer]
mmu_man has joined #haiku
mmu_man has quit [Ping timeout: 480 seconds]
mmu_man has joined #haiku
DKnoto_W has quit [Quit: Leaving]
DKnoto has joined #haiku
nephele_xmpp has joined #haiku
nephele_xmpp has left #haiku [#haiku]
nephele has joined #haiku
zard has joined #haiku
<nephele>
Hi there
<zard>
Hello!
<nephele>
hi zard :D I got help and now my system is properly booting in 4k from a ssd that now only has haiku on it
<nephele>
only problem now is getting clang to compile webkit xD
<zard>
Ooh, going to try to get clang to compile WebKit? I wonder how that works for you
<nephele>
apparently haikuports already does this
<nephele>
so it's bound to work with some (smaller) fixes
<zard>
Ok, interesting
<Begasus>
hi nephele zard
<Begasus>
atleast you got lucky nephele, setting it up on BM :)
<nephele>
Begasus: it's kind of cool really, severall small fixes by severall people made me be able to set it up xD
<nephele>
also can set my screen to 120 (or 140) hz refresh rate, but i dunno how to test this actually works :(
<Begasus>
don't go overboard :)
<nephele>
My screen is 240hz Begasus :)
<Begasus>
just saying, small things can lead to ... :)
<nephele>
higher refresh rate is less latency added by the monitor ideally, and less latency leads to less errors in operating a computer
<Begasus>
if primary target is to have a build system, I guess that refresh rate doesn't realy matter?
<nephele>
in that case the resolution would also not matter, but i'm programming on this machine directly
<nephele>
and probably phase out my i3 small one and move my stuff there
<Begasus>
in any way progress then :D
<nephele>
zard: seems clang comnplains about inclusions of c++/bitset
<nephele>
if it is used
<nephele>
but seems that maybe i can get around that? :g
<nephele>
haha, build failed
<nephele>
waddlesplash: no member fromUTF8 in StringView, i guess that answers your question :D
<nephele>
only 3,6K targets to go
<phschafft>
Is there anyone specifically into Haiku and perl?
imrahil has joined #haiku
<Begasus[m]>
I think jmairboeck has some good knowlidge?
<nephele>
phschafft: i have a signbed perl6 book, well before they renamed it to raku
<phschafft>
hm.
<phschafft>
nephele: fun. :)
<nephele>
one of those books i wanna pick up again, some things in raku are really really neat, like how it deals with unicode amazed me
<phschafft>
I'm more on the Perl 5 side. but Perl 5 and Raku kind of pull each other up.
<nephele>
I think renaming it from perl6 to raku was a smart move
<phschafft>
maybe yes.
<phschafft>
so, my new module, just tested, kind of worked out of the box in Haiku.
<nephele>
well, it's one of those "it's similar but not the same" situations, i think saying they are in a similar language family makes more sense than saying "yes its the next version"
<phschafft>
the default example complains about two missing optional dependencies. that is something that can surely be fixed.
<phschafft>
nephele: it also allowed Perl to continue growing as Perl. without being 'the old thing'.
<nephele>
yeah
<phschafft>
it also can't find any filesystems. not really sure why. but that is a completly different question.
<phschafft>
perldoc however doesn't work. it complains about no write permission to write in ~
<phschafft>
so I guess there is something else going on there.
<Begasus[m]>
stupid Debian, says fdisk is installed, but launching (or trying to) says it can't find it
<scanty>
Begasus[m], /usr/sbin/fdisk
<phschafft>
Begasus[m]: mind your path.
<scanty>
and that's it until later, going to work.
<phschafft>
You don't work here?
* phschafft
looks around in a 'what is this place anyway'-fashion.
<Begasus[m]>
err ... does that need a path?
<Begasus[m]>
f*n annoying login screens ...
<Begasus[m]>
had this the other day, live session for Neon, screen locks and asks for a pswd, which one!! :P
<phschafft>
Begasus[m]: on classic POSIX systems system admin tools are not your path if you're not root.
<phschafft>
nothing to do with Debian.
<Begasus[m]>
phschafft already had switched to root with "su"
<Begasus[m]>
anyway, worked now, dd'd some zero's to the USB drive
<phschafft>
Begasus[m]: you need to use 'su -' to get a super user env.
<phschafft>
'su' alone will try to keep as much from your env as possible.
<phschafft>
which includes environ and cwd.
<Begasus[m]>
ah, didn't know that (too long since I last used it) thanks!
<phschafft>
generally avoid using 'su' and always do 'su -'.
<phschafft>
(I don't say all that is logical or valid in 2024, just telling how it is and that it has been that way for... very long)
<Begasus[m]>
Debian probably wont last that long on that laptop, it's just one linux install that doesn't s*ck as much as the other ones on older hw :)
_-Caleb-_ has joined #haiku
<nephele>
can't one do a git init on fat32?
<phschafft>
nephele: what's the error?
<Begasus[m]>
no idea, NTFS here on Windows where a git repo also is, others BFS
<phschafft>
Begasus[m]: we use Debian a lot specifically because it just works whatever setup. they call it the universal OS, and that is kind of what it is.
<phschafft>
so it's at least a good default go-to.
<phschafft>
nephele: on which OS?
<Begasus[m]>
yeah, had been using it years ago, out of practice I guess phschafft :D
<nephele>
Haiku
<phschafft>
nephele: hm.
<phschafft>
nephele: mkdir .git
<phschafft>
does that result in the same error?
<nephele>
yep
<waddlesplash>
it may want FS features FAT does not support
<phschafft>
nephele: then you can only go with a bare repo.
<phschafft>
I wonder if that check for the filename is correct however. maybe it's a bug in the fs driver.
<nephele>
already had enough problems that "boot" can't be renamed to "BOOT" on fat32 because the "file already exists"
<nephele>
and almost every file not beeing able to be opened by the gui, telling me it was "mistakenly marked as executable", the dialog to fix this doing nothing, and files never beeing identified
<phschafft>
I mean that is correct. boot and BOOT are the same files on fat.
<nephele>
phschafft: yeah, but it should let me rename it. Even if it is the same file the casing is preserved
<waddlesplash>
i thought that was fixed recently
<phschafft>
preserving case is a mount option. not a filesystem feature.
<phschafft>
so that is something which might depend on @factors.
<nephele>
If it shows me two different cases clearly it has them stored somehow
<phschafft>
nephele: my linux allows for a directory called '.git'.
<phschafft>
however maybe it uses long file name extention or something for that.
<phschafft>
nephele: it has the filename as bytes but the specs are that the case does not matter.
<phschafft>
so that is purly for display IF at all. many implementations also normalise it to one of the cases.
<nephele>
then it should not matter to haiku either if i rename it to another case ;)
<phschafft>
nephele: then it must also not fail if you do 'mv bla bla' on some other filesystem.
<phschafft>
pick one side.
<nephele>
sure it should...
<nephele>
I really don't see where your problem is with this, one time there is a change, in the other case there is not
<nephele>
Even if the change is "Only visual"
<phschafft>
and may not be even that.
<nephele>
nice. my refind theme (with sources) is 416KB
<phschafft>
I mean if the case doesn't matter the driver is free to normalise it on write as well.
<phschafft>
nephele: generally speaking mv will fail on two times the same file.
<phschafft>
let's see what POSIX has to say!
<nephele>
haiku does not normalize the case, so that is a moot point.
<phschafft>
If the old argument and the new argument resolve to either the same existing directory entry or different directory entries for the same existing file, rename() shall return successfully and perform no other action.
<phschafft>
as vor mv(1) POSIX is more complex. however it allows it to be a no-op.
<phschafft>
a.
<phschafft>
No change is made to source_file, no error occurs, and no diagnostic is issued.
<nephele>
hmm, did my mail send the mail or did it not
<Begasus[m]>
that's the question* :)
<nephele>
Haiku Inc: Haiku builds games that train
<nephele>
Haiku stands at the forefront of cybersecurity upskilling, leveraging video games to immerse you in a flow state for accelerated, enduring learning.
<nephele>
of ffs
<Begasus[m]>
still nogo ... "Boot volume is not valid" :'(
<nephele>
i got that yesterday when the installer did not change from "linux data" type, but used the bfs partition on it as is :D
<Begasus[m]>
would love to get to the stage of Installed :)
<Begasus[m]>
s/Installed/Installer
Peppersawce has joined #haiku
Begasus has quit [Quit: Vision[]: i've been blurred!]
bbjimmy has quit [Quit: Vision[]: i've been blurred!]
<nephele>
zard: nope, not getting past that bitset error
<zard>
heh
HaikuUser has joined #haiku
bbjimmy has joined #haiku
<HaikuUser>
/boot/system/develop/tools/lib/gcc/x86_64-unknown-haiku/13.3.0/../../../gcc/x86_64-unknown-haiku/13.3.0/include/c++/bitset:1344:7: error: constexpr function's return type 'std::basic_string<char, std::char_traits<char>, std::allocator<char>>' is not a literal type
<zard>
O.O
<phschafft>
nephele: which files is +x set on?
<nephele>
phschafft: context?
<phschafft>
fat mount.
<phschafft>
13:42 < nephele> and almost every file not beeing able to be opened by the gui, telling me it was "mistakenly marked as executable", the dialog to fix this doing nothing, and
<phschafft>
files never beeing identified
<nephele>
I've not checked. Only observed the gui complaining
<phschafft>
ah.
<phschafft>
I have seen several levels of smartness in drivers handling that.
bbjimmy has quit [Quit: Vision[]: i've been blurred!]
<nephele>
seems to be new-> text file mistakenly creating stuff marked executable
<nephele>
and for some reason the dialog to fix this doesn't work
HaikuUser has quit [Quit: Vision[]: i've been blurred!]
<nephele>
zard: yeah, no clue what to do about that error *shrug
<nephele>
have some slightly different ones too, but also in that file
<nephele>
so... i guess wasting another day just compiling webkit, and this time again with gcc .-.
<zard>
Ah, yes. We all love WebKit's compilation time \s
<phschafft>
I mean setting the file modes will generally not work with fat (some cases may, also depending on how smart the driver is).
<phschafft>
but altering the execution bits on a file has no meaning to fat.
<zard>
I wonder, why did you want to use clang over gcc?
<nephele>
waddlesplash mentioned the memory usage was significantly lower, and gcc would scrape my 32GB of ram and my system would become unresponsive
<nephele>
with disabeling aslr and clang anyhow
<nephele>
and he was correct, the memory usage was about 20GB usually then
<zard>
Tip: lower the thread count till it uses just as much memory as you have available
<zard>
Probably already knew that though :)
<nephele>
I know that option, but i got a 12 core cpu exactly to compile webkit xD
<nephele>
well, 6 core... 12 threads
<zard>
Exactly the same as my computer :D
<nephele>
so should i get another RAM stick? :P
<zard>
Never can have enough these days!
<nephele>
my motivation is a bit gone, now that the build failed yet again
* Begasus[m]
knows the feeling ...
linuxmaster has quit [Ping timeout: 480 seconds]
DKnoto_W has joined #haiku
DKnoto has quit [Ping timeout: 480 seconds]
<waddlesplash>
nephele: what happened with clang?
<waddlesplash>
clearly it works, since I built an entire WebKit with it
<nephele>
It fails compiling headers included from gcc13
<nephele>
> <HaikuUser> /boot/system/develop/tools/lib/gcc/x86_64-unknown-haiku/13.3.0/../../../gcc/x86_64-unknown-haiku/13.3.0/include/c++/bitset:1344:7: error: constexpr function's return type 'std::basic_string<char, std::char_traits<char>, std::allocator<char>>' is not a literal type
<nephele>
maybe the haikuports version pulls in an older gcc...
<nekobot>
[haiku/haiku] 1db0961121bd - kernel/vm: Remove an obsolete comment.
<nekobot>
[haiku/haiku] 5e451463d00f - kernel/vm: Add assertion for a TODO comment in mark_page_range_in_use.
<waddlesplash>
Begasus[m]: what about chainloading from rEFInd?
<Begasus[m]>
the other laptop is even a lot older and has the same issue
<Begasus[m]>
never worked with rEFInd
<Begasus[m]>
eg never used it
<Begasus[m]>
and even then, I can select EFI boot from the boot menu, select the one from the USB thumbdrive ... still the same
HaikuUser has quit [Quit: Vision[]: i've been blurred!]
diver has quit [Ping timeout: 480 seconds]
<Peppersawce>
Running Gog installers via explorer via BoxedWine is pretty cursed
<Peppersawce>
At least it works :D
<nephele>
depending on what you need maybe that innoextract script makes more sense
<nephele>
"Inno Setup is a tool to create installers for Microsoft Windows applications. innoextract allows to extract such installers without running the actual setup executable under Windows or using Wine."
mmu_man has quit [Ping timeout: 480 seconds]
<Peppersawce>
Huh, might try that
<nephele>
might make sense if you intend to run a game in scummVM on Haiku or something, and have a gog license
<Monni>
80000004 would be B_BAD_TYPE I think...
<Peppersawce>
actual wine is screwed up
<Peppersawce>
gog technically has a downloader script
<nephele>
it's just not a finished port *shrug*
<Peppersawce>
But the libraries it needs do not work
<Peppersawce>
I'm hacking one of the many solutions together but maybe Inno works better I should try
<nephele>
now that you mention it i'm gonna play some blade runner if i can get it to run
<nephele>
*** Failed to find a match for "devel:libboost_filesystem>=1.37.0": Name not found
<nephele>
>:(
<nephele>
why not
<Begasus[m]>
lol
<Begasus[m]>
we don't have a 1.37.0 version
<nephele>
so what?
<nephele>
it sais >=
<nephele>
not ==
<Begasus[m]>
make that >= 1.83
xet7 has quit [Remote host closed the connection]
<nephele>
that better not work begasus >:(
<Begasus[m]>
well, innoextract needs a rebuild, or boost1.85 needs one with ICU74
<nephele>
It works. Great. That means haikuporter is broken
<Begasus[m]>
fixing boost1.85 should fix it for both innoextract and VCMI
<nephele>
not happy about that
<Begasus[m]>
patches welcome (by now I got a pretty good idea how it works)
<Monni>
Haikuporter being broken is nothing new... It hasn't had much TLC in 10 years...
ks has joined #haiku
<nephele>
I don't want to write python code
<Begasus[m]>
there have been a few nice improvements not that long ago
<nephele>
and i question the wisdom of making a build system based on shellscripts
<ks>
does haiku can be running on arm device?
<Monni>
Need to be drunk enough to even touch Python code...
ks has quit [Remote host closed the connection]
<nephele>
ks: there is no working arm or arm64 port, iirc arm64 can boot in qemu somewhat
<PulkoMandy>
nephele: The definition of >= in haikuporter is not what you expect
ks has joined #haiku
<nephele>
uh oh... what is it instead?
<PulkoMandy>
It was still better to use characters you can type on any keyboard than some obscure mathematical operator
<phschafft>
ok, code updated. all not nice but hey.
<PulkoMandy>
If I remember correctly, >= 1.35 means 1.35.anything
ks has quit [Remote host closed the connection]
<nephele>
wait... so no way to declare dependencies permissively?
<PulkoMandy>
I think this is documented in haikuporter wiki or in package kit internals documentation
<PulkoMandy>
you can do >= 1 then it is 1.anything
<PulkoMandy>
But otherwise no, not really, because this was designed mainly for soname comparisons, and if your soname doesn't match, the library won't be found by runtime loader
<nephele>
Yes, but. If an application can be compiled with boost 1.37 up to 1.85 or whatever is a different problem to wether it *was* compiled with a specific version
<Begasus[m]>
using >= 1.37 as in expecting the first in line would be using 1.69 .... jikes no!
<PulkoMandy>
yes. But it is also something we don't really care about
<nephele>
I care about haikuporter beeing robus. It does not make sense if it can't deal with these constraints on it's own
<PulkoMandy>
Really we want to try to get things down to only one boost version if possible
<nephele>
Ideally sure. But that isn't helped when haikuporter tries to force me to use a higher boost version than i have already installed ;)
<PulkoMandy>
So we don't need the info on old compatibilities, just "was this built with the latest version, or not yet?"
<Begasus[m]>
have been cleaning enough already on 1.69.0, probably should be good to disable by now
<nephele>
there is source compat for this with older boost versions, that info is lost when added to a recipe
<PulkoMandy>
Yes, because it should be irrelevant. If I moved something to a newer version too early, that is my mistake
<Begasus[m]>
iirc it worked ok PulkoMandy?
<nephele>
I dunno. How do you move something too early? I mean, you couldn't have known beforehand that icu75 would be disabled?
<PulkoMandy>
there can be a "main version in use" of the recipe (as there is for python 3.10 for example)
<PulkoMandy>
And we can try to stick to that until we are ready to deploy a new one widely
<nephele>
sounds like another situation that would be helped with testing branches...
<PulkoMandy>
But that is alo information not really tracked anywhere. I don't know the best way to track it, could be in the recipes, or could be a wiki page at haikuports for example
<PulkoMandy>
Yes, some branch discipline could work too
<Begasus[m]>
those base packages have no conflicts like with the python packages
<Begasus[m]>
you can still have older ICU versions installed, just didn't/don't get why ICU75 was a bad thing to have around
<PulkoMandy>
You can have multiple versions installed, but it's nice to only have one at least during haiku releases
<Begasus[m]>
but it is as it is now, not changing anything there
<PulkoMandy>
Otherwise people complain that haiku is bloated
<Begasus[m]>
there was PulkoMandy ICU74 has been default in the nightlies for quite some time
<nephele>
haiku is bloated, we have webkit three times ! :D
<PulkoMandy>
So... more politics and marketing than technical reasons really
<Begasus[m]>
you don't have to ship ICU75 for Haiku, there is no need for that
<waddlesplash>
the problem is that if we let these dependencies around, then they creep into things that Haiku itself depends on
<Monni>
Everyone wants their favourite application ported to Haiku... That's kinda dictionary definition of bloated...
<waddlesplash>
there are more packages in the base install that depends on ICU than just Haiku itself
<waddlesplash>
we want them all to depend on one ICU version not multiple
<Begasus[m]>
building Haiku depends on a specific package afaik?
<PulkoMandy>
And speaking of webkit: waddlesplash did build the latest release with clang, but I think not the latest commit of the current branch. So that may need more fixes. For now I'm personally focusing on gcc because I know that better and can't do everything :)
<PulkoMandy>
Begasus[m]: Yes, baut freetype and harfbuzz may depend on another, for example
<nephele>
that's fine :) I'm going to try and use gcc now too... but those memory savings are really something
<PulkoMandy>
And haiku depends on those
<nephele>
maybe haiku inc will send me a second ram stick xD
<PulkoMandy>
Yes, I'm building with -j4 to not run out of ram currently
<Peppersawce>
For one I'm confident you developer guys are smart enough to sort the webkit situation out eventually :)
<Peppersawce>
We're rooting for you :)
<Begasus[m]>
if so, make the "require" ICU74, that shouldn't be to hard?
<nephele>
Peppersawce: your name looks too much like PulkoMandy from a glance xD
<Peppersawce>
lol :D
<PulkoMandy>
No relation!
<Begasus[m]>
anyway, enough on the subject
<nephele>
PulkoMandy: I'll try with -j10 or -j8 next i suppose
<nephele>
though this board can be upgraded if needed. but i have to replace that damn cpu cooler if i do
<nephele>
having clang do it would be neater eventually
<PulkoMandy>
Begasus[m]: Yes, shouldn't be hard but it's not done currently. Maybe a comment with a reminder in all recipes used by Haiku would be enough
<nephele>
Anyhow, i'm focusing on the webkit2 branch and trying to implement stuff we need
<PulkoMandy>
Nice :)
<nephele>
seems mousewheel events, keyboard events, and then trying to see if i can fix the dark mode, fix native rendering of some stuff (no popups?) etc
<PulkoMandy>
I'll not be doing much Haiku next month I think. I have planned a 900km bike trip over two weeks and I also have two talks to prepare for Capitole du Libre (one about Haiku and one about a smalltalk interpreter for Amstrad/Schneider CPC computers)
<PulkoMandy>
After that hopefully it will be back to "normal" (but I always say that and it never happens)
<nephele>
nice! a bike trip sounds fun :D
<gordonjcp>
PulkoMandy: awesome
<gordonjcp>
PulkoMandy: where's the bike trip through
<gordonjcp>
?
<PulkoMandy>
Going from Toulouse to some small village in Normandy for an Amstrad CPC user meetup
<PulkoMandy>
So I had the probably stupid idea to bike that
<botifico>
[haikuports/haikuports] korli 241d99d - qt5: bump version
<PulkoMandy>
zdykstra: There will be a few stops along the way and taking a more bike-friendly route
OrangeBomb has quit [Quit: Slacking off]
<PulkoMandy>
Probably along the atlantic coast
<nephele>
PulkoMandy: in relation to bloat and libraries, on one hand i can understand the want to only have to have one version so the system is "sleek" and "modern". but on the other I am hugely annoyed that the constant upgrading of frameworks like kde's libs keeps throwing projects under the bus if they don't have a maintainer
<nephele>
I would really like it, in theory, if i could just install qt3, qt4 apps and such. But i guess that just isn't in the cards either with our workload for maintenance :)
<Monni>
Qt3 brings memories...
<zdykstra>
PulkoMandy: that sounds like a fun trip. I assume you've done distances like that before?
HaikuUser has joined #haiku
DKnoto_W has quit [Quit: Leaving]
DKnoto has joined #haiku
HaikuUser has quit []
HaikuUser has joined #haiku
HaikuUser has quit []
HaikuUser has joined #haiku
<PulkoMandy>
Not at all. I sometimes do about 50km a day on weekends. We'll see what happens :)
<PulkoMandy>
(But I can give up at anytime and take a train)
<Peppersawce>
Monni, while building lgogdownloader cmake complains it can't find git here
<Peppersawce>
And had the binary crash when man invoked it for info both times
<nekobot>
[haiku/haiku] d2b9dcde3129 - kernel/vm: Only print "not implemented" messages in VMTranslationMap under KDEBUG.
<SArpnt[m]>
still can't boot haiku off a hard drive
<SArpnt[m]>
tested on a different computer too now
_-Caleb-_ has joined #haiku
HaikuUser has joined #haiku
HaikuUser has quit []
<SArpnt[m]>
i cloned the entire working bootable flash to the hard drive and it just doesn't boot
<MonniTheCat>
That obviously won't work...
<SArpnt[m]>
why not? as far as i know information about mbr boot is stored entirely on the disk and dd should clone it
<SArpnt[m]>
even if that didn't work surely across two different computers and two different drives either the same image i flashed or the haiku installer should work
<MonniTheCat>
Flash drives are like CD-ROMs or floppies... there is extra data that needs to be stripped...
<SArpnt[m]>
i haven't heard that before
<SArpnt[m]>
i mean i'm sure it does store extra data somewhere but is that exposed anywhere in the linux /dev/* block device?
Atomozero has quit [Quit: Vision[]: i've been blurred!]
<MonniTheCat>
there is tools in Linux to see where the extra data is and where the real partitions start...
<SArpnt[m]>
i know the partition image certainly doesn't include anything extra
<SArpnt[m]>
and that didn't work on the hard drive
<SArpnt[m]>
i don't think you were here last time i should probably go over everything i've tried
<SArpnt[m]>
i haven't been able to install haiku for days and booting it was difficult
<SArpnt[m]>
flashing the iso with dd and balenaetcher doesn't boot
<SArpnt[m]>
* sd card and i still have that and can boot off it and use it fine
Atomozero has joined #haiku
jmairboeck has quit [Quit: Konversation terminated!]
<SArpnt[m]>
but if i try the same thing on a spinning hard drive it doesn't work
<MonniTheCat>
You can use the SD card to create one big BFS parition on the real hard disk and make it bootable...
<SArpnt[m]>
i have tried this with two hard drives and two computers
<SArpnt[m]>
the haiku installer used according to the instructions also doesn't work
<SArpnt[m]>
just now copying the entire sd card block device to the hard drive doesn't work
<SArpnt[m]>
MonniTheCat: how
<SArpnt[m]>
like i said the haiku installer doesn't work
<SArpnt[m]>
on both computers it just goes to a blinking underscore cursor and nothing else
<MonniTheCat>
as long as you did run makebootable before restarting after running the Installer, it should boot... If it doesn't, then the bootloader is broken...
mmu_man has joined #haiku
<SArpnt[m]>
i did run makebootable and it doesn't boot
<SArpnt[m]>
it worked fine for the sd card
<MonniTheCat>
Then only solution is to use Linux live CD and install GRUB and chainload Haiku using it...
<SArpnt[m]>
why would the bootloader be broken on a hard drive
<SArpnt[m]>
does it not have a driver for it or something?
<waddlesplash>
if it didn't, you'd get a "no boot partitions" message
<MonniTheCat>
I'm not sure as it works fine on my PC I bought in 2019...
<waddlesplash>
I would recommend just booting from the SD card and installing to the hard drive
<SArpnt[m]>
waddlesplash: i said many many times that doesn't work
<waddlesplash>
okay, what happens?
<waddlesplash>
"blinking cursor" usually means we have not even got into the Haiku loader at all
<SArpnt[m]>
blinking underscore nothing else
<SArpnt[m]>
with mbr boot
<waddlesplash>
if the system supports EFI you can try that
Xanxair has joined #haiku
<SArpnt[m]>
the system i'm trying to get it on doesn't support efi
<waddlesplash>
ok. very odd that it boots from the SD card but not the HDD
<waddlesplash>
but "a blinking cursor" is almost certainly the BIOS itself
<waddlesplash>
and we don't even get to the Haiku bootloader
<SArpnt[m]>
but for testing i've tried two computers and two hard drives and every installation method i could find
<waddlesplash>
well, that's strange, because most systems I just flash the standard ISO and it works fine
<waddlesplash>
old, new, etc.
<MonniTheCat>
With MBR, only thing that stops it from seeing Haiku bootloader, if the BFS partition isn't set to active and bootloader is in the VBR and nor MBR...
<SArpnt[m]>
the bfs partition is active and the linux makebootabletiny tool says everything is fine
<waddlesplash>
what is "makebootabletiny"?
<waddlesplash>
you should not need any "makebootable" if you are using an official image build
<MonniTheCat>
Then only solution is to use null modem cable and read the boot logs that way...
<waddlesplash>
in 99% of cases you never need to run makebootable
<waddlesplash>
I can't recall the last time I ran it
<waddlesplash>
because 1. default images have this already set, and 2. Installer does it for you
<waddlesplash>
there are very few circumstances where you need to invoke it manually
<SArpnt[m]>
if you didn't see earlier flashing the iso with balenaetcher also didn't work
Xanxair has quit []
<SArpnt[m]>
i had to pull an image off it and flash that to an sd card partition instead
<waddlesplash>
this is what it looks like to have BFS take up the entire disk
<waddlesplash>
no partition table
<SArpnt[m]>
yeah i never had that
<waddlesplash>
ok
<waddlesplash>
does it say "Intel Partition Map" then?
<waddlesplash>
you can also try GUID partitioning
<SArpnt[m]>
i don't think my old laptop will take gpt
<SArpnt[m]>
and that's what i want to actually install on
<SArpnt[m]>
even if it does last i remember gpt+bios boot is very weird and haiku has no instructions for it
<waddlesplash>
the Haiku bootloader should tolerate GPT
<waddlesplash>
though admittedly this is indeed an uncommon setup
<SArpnt[m]>
and if i dualboot later then all the other oses will definetly struggle
<waddlesplash>
I guess I just don't know what could be going wrong here
<waddlesplash>
you can try installing the Haiku Boot Manager
tqh has quit [Quit: Leaving]
<waddlesplash>
this is a basic boot partition selector
<SArpnt[m]>
i'm not dualbooting now because i couldn't figure out how to get windows 2000 installed (it's the highest number windows) and i haven't found any 32 bit linux distro i like
<waddlesplash>
so, maybe try installing that and seeing if it fixes things
<waddlesplash>
just run BootManager application and select the drive to install on
<waddlesplash>
should work to dualboot with anything that supports basic chainloading
<SArpnt[m]>
ok haiku booted and i just checked
<SArpnt[m]>
it says intel partition map wayy off to the right
<waddlesplash>
OK, good
<waddlesplash>
try installing BootManager on this drive then
<SArpnt[m]>
run the regular installer then bootmanager afterwards?
<waddlesplash>
if you already ran the regular installer you don't need to run it again
<waddlesplash>
just run BootManager and install it on the drive
<SArpnt[m]>
i can't find bootmanager
imrahil has left #haiku [Disconnected: Hibernating too long]
<coolcoder613>
waddlesplash: Are those memory saving changes in Beta5 or only in the nightlies?
<MonniTheCat>
... /boot/system/apps/BootManager
<SArpnt[m]>
haiku boot manager worked!
<MonniTheCat>
So the MBR was corrupt...
Atomozero has quit [Quit: Vision[]: i've been blurred!]
<SArpnt[m]>
how would the mbr be corrupted
<SArpnt[m]>
i remade that partition table many many times
<SArpnt[m]>
* many times in gparted
<SArpnt[m]>
and i tried two hard drives
Atomozero has joined #haiku
<MonniTheCat>
I don't use gparted, I only use fdisk on Linux...
<SArpnt[m]>
i use both for different things
HaikuUser has joined #haiku
<SArpnt[m]>
gparted can resize most partitions and their filesystems
ToyDragon has quit [Ping timeout: 480 seconds]
<waddlesplash>
coolcoder613: nightlies. Some are a little dangerous so let's see what happens with them
<MonniTheCat>
Only thing I know about MBR is that it has own signature at the end of the sector...
<MonniTheCat>
Never looked at the "default" bootloader that is supposed to chainload the active partition...
HaikuUser has quit [Read error: Connection reset by peer]
<coolcoder613>
Haiku boots in 128MB...
<coolcoder613>
Takes longer to press the "Try Haiku" button than it does to boot
<coolcoder613>
Just got the empty blue screen with a mouse cursor now... time to wait
Atomozero has quit [Quit: Vision[]: i've been blurred!]
<waddlesplash>
nice!
<waddlesplash>
there are other changes one would make if you really wanted to use a system in that small of memory
<waddlesplash>
a bunch of options to turn off etc.
<coolcoder613>
Still the same... I think I'm going to try increase the RAM
<coolcoder613>
Trying with 136m now
<coolcoder613>
Runs much better
<coolcoder613>
"An error was encountered and the installation was not completed: Error: Out of memory"
<PulkoMandy>
SArpnt[m]: I don't know what loader code gparted puts in the mbr. If you recreate the partition table from DriveSetup it should put our own bootcode there, or you can also do this with the writembr command
<PulkoMandy>
Otherwise we don't touch that and leave whatever was there (in case you have grub or some other thing set up)
<PulkoMandy>
Maybe drivesetup or installer should warn about "unknown" boot code but there are dozen of variations of basically the same thing (from the various bsds, from different versions of windows, etc)
<SArpnt[m]>
i already have the system running
<SArpnt[m]>
also again doing the exact same thing on the flash sd card booted while the drive didn't
<SArpnt[m]>
so i don't think it's a matter of something not being set up
<SArpnt[m]>
there was something involved that worked on the sd card but didn't work on the drive
<PulkoMandy>
There are some weird things happening, for example the bios could be ignoring the master boot record on external devices and booting the active partition directly
<PulkoMandy>
But not on internal hdd
<SArpnt[m]>
is this common?
<SArpnt[m]>
because this was on two computers
<MonniTheCat>
Most USB sticks use either floppy or CD emulation like I already said...
<PulkoMandy>
Yes, because usb drives often have a mbr that just prints "failed to load operating system, press a key to reboot" or something like that
<PulkoMandy>
So that ends up preventing to boot the machine if you have an usb key connected
<PulkoMandy>
MonniTheCat: Usb sticks don't do that. They are just usb mass storage, there is no need to make it more complicated...
<SArpnt[m]>
ah makebootabletiny does write to a partition specific boot record not the disk record
<SArpnt[m]>
that's probably it then
<MonniTheCat>
PulkoMandy: Even my motherboard BIOS updater use floppy emulation on USB stick... This is common on PCs with no physical floppy drive...
<SArpnt[m]>
makebootable should probably be documented better and i think the installer should warn if it didn't fully install a bootloader
_-Caleb-_ has left #haiku [#haiku]
<SArpnt[m]>
i also find it odd it's not mentioned anywhere on here?
<SArpnt[m]>
a freshly formatted disk isn't going to come with an mbr bootloader
<SArpnt[m]>
and that page is for completely skipping the usb drive
<PulkoMandy>
Makebootable is never needed. I don't understano why people keep finding it and trying to use it
<SArpnt[m]>
because it's in the documentation right there
<SArpnt[m]>
and i'm pretty sure multiple people told me to use it
<PulkoMandy>
they are all wrong :(
<SArpnt[m]>
this doesn't even look like a wiki it's just the website
mmu_man has quit [Ping timeout: 480 seconds]
mmu_man has joined #haiku
<MonniTheCat>
makebootable is only useful with old pre-UEFI PCs when moving raw BFS filesystems from one disk to another or to another partition on same disk...
<SArpnt[m]>
this is a pre-uefi pc
MonniTheKitten has joined #haiku
<waddlesplash>
PulkoMandy: well, it's *sometimes* needed. But that's so rare it's basically never