p dig

oops

eh Harzilein where u playing pirates

i know that command p fish

p love aye

:)

aye makes p love a no-op though

where u playing

xeromem

and we happen to need three more players for next season

so feel free to join us on rising sun

where is that

irc.xeromem.com

night seamonkey

ok tell me more later i gtg chat with u again

tomman: what is your UA? OS?

tomman: I think the only thing I read on his post suggesting a problem was him blocking older versions of Windows NT - I mean, I'm guessing the Firefox part of the UA string won't be triggering yet, but if it does, well, then I guess that needs fixing

tobin: if *all* distros are on wayland by 2027, I won't be using Linux-based systems by 2027, I guess.

0509|21:13:31 < Tekk> Most of the DE screensavers are shit forks of xscreensaver <-- one issue he repeatedly mentions is that a lot of them seem to have worse designs for screen locking, sounding like there are two ways to lock at least an X11 screen: xscreensaver or vlock -a.

things like error conditions or corner cases leading to unlocked screens in other screensavers with a screen locking feature

tomman: if GCC is outputting wrong code, that surely does look like a bug, my only guess would really be that someone somewhere writing code for GCC read "Pentium" and thought "oh it means IA-32", forgot how long IA-32 spans, and added that...

0510|14:49:15 < UsL> yes, we desperately need a anti-enshitification movement <-- "broken by design" as a slogan has existed for years, so I guess there'll at least be the fight against DRM; also, anything on that route would hopefully be mentioned by Cory Doctorow at some point

tomman: I think it's imgur that has opengraph or twitter meta tags which don't work for me, I need to see if some UA makes these load as images...

njsg: yikes, i retain some remnant of "framebuffers are optional" from my formative years.

njsg: the wayland thing has secondary effects for sure. but if it's all broken by 2027, we can work it out.

I think the main thing might be having enough people gathering around X11 to start a fork of Xorg. Who else has been contributing changes besides NetBSD?

i'd hope openbsd too?

edbrowse (3.8.12-1) UNRELEASED; urgency=medium

* new upstream version

-- Sebastian Humenda <shumenda⊙gd> Mon, 05 May 2025 09:13:56 +0200

:)

nice

wayland suppor is probabyl needed later. I have no time and I am mostly on windows anyway. Stuff resides mostly in widget. Can provaide a patch queue if someone wants to start doing backports.

patch queue => you'll keep using hg?

yes.

git is unsuiteable fdor doing backports. Would take forever to rebase stuff.

"git is unsuiteable", seamonkey dev claims :3

If were not for the mq extension it would not matter. Also Tortoise Hg is such a convient tool. They can put phab and all the stuff where the sun does not shine.

meanwhile, the latest in barrier of entry is Anubis™.

Kind of makes CloudFlare look acceptable, IMHO.

I'll take Anubis over Clownflare, but some sites are already starting to crank it up to "System Requirements: overclocked Core i9/Threadripper"

gitlab wip updated

njsg: JWZs is still messing with his blocklists, but at least he dropped the latest FF ESR UA from his "naughty" list

he hasn't also blocked my "special" UA yet, but I guess it doesn't smell enough to a bout :D

--bot

And imgur really really REALLY doesn't want you watching images without a valid referer tag, it seems

...is there an addon compatible with SeaMonkey that can arbitrarily inject/manipulate HTTP headers?

njsg: anubis is fine as much as it "promises" any reasonable competent reader can make sense of the code.

reasonably*

and indeed i can get challenges

hello seamonkeys :)

can somone offer me some nice links about what i all can do with that composer of seamonkey, please :) ?

have you read the help for it in-application?

just a bit

but maybe somone has some experiences with it here, i thought and might be able to tell me abou some services that i could use for publishing with it??

i tried with a 'write.as' account but that seems not to work, cause it won't let you except you copy paste your edited text there

somthing different: i have trouble to log in to a webpage i can login to with firefox.. do i need to change preferences?

grobi: depends on the page, settings-wise, you can always try playing with the three built-in user-agent strings, if that does not work, you can try setting different UA strings manually (if that works, you can define a per-domain UA override for that domain); now it's possible the problem is it requires features not implemented yet. Decent-ish pages will at least print some error to the JS console that

you can use to track down what's missing (better pages have error handling that allows it to work, lookin' at you Cloudflare and Anubis)

...

well, then.

if they are using chatzilla restarting closes irc remember

njsg:

grobi: with a suitable machine and enough time, you can always try several Firefox versions to see when does it start working (now the site might somehow hit some threshold or might just need a cookie set to work, so this might still not uncover what's missing. I guess binary search is appropriate here.

grobi: once you narrow down to the Firefox release where it starts working, you can try to figure out which of the changes is behind it, or at least mention that it works starting with that version

nsITobin: hopefully there's always logbot too

grobi: besides JS features, it could also be related to HTTP headers (/me looks at Cloudflare's self-DDoS-because-you-don't-have-Origin:)

njsg, thank you for your answer, unluckily i just found out, that seamonkey is not the right browser for the distro i'm on ..

... oh, why?

.. right now. It just take way too moch space in my home folder

i'm on porteus savefile and had to notice, that even 1,5GB added to my savefile was enough to serve space for seamonkey

There is also the WIP branch, for which there might be builds (are these on archive.seamonkey-project.org?), but whether these can actually have the missing features might be a shot in the dark, as there are also other features temporarily missing in that branch

was not enough^

you mean profile or the size of the program files?

for the former it might be possible to disable features like disk cache, but it will still be a though situation if you hit the space limit. places db might grow but for that one you could set self-deletion of older history, I'm not sure if that helps much but should at least prevent excessive growth (actually, wasn't there a newer setting that adapted to existing space? how's that one?)

seamonkey itself i installed and activated here as an xzm module like all other apps here on porteus-linux is not that big

but it installe a tremendous amoount of config? files in my homefolder ..

don't know why..

re: composer I intend to some day see how the publish code currently works, I haven't used "Publish" there, at least not this decade. Maybe someone else knows how it works

grobi: how big is too big there?

i will install seamonkey to another distro soon and will try it

njsg, it took ~1,4GB in my homefolder which is not normal i guess??

so no wonder it just runs bad on a live-system with a now 3GB save-file ..

grobi: and this is just profile data, not the program files? All under ~/.mozilla or is this ~ overall?

grobi: I'm thinking the likeliest source for that would perhaps be cache, but nowadays isn't that in ~/.cache?

all under ~/.mozilla

i did not even look at the cache yet lol

maybe also message storage if you're using M&N and saving for offline usage. But here really it'd be appropriate to check the subdirectories and files and see if anything stands out

if .mozilla is 1.4 GiB, unless you have dozens of profiles you've used for some time (probably not the case, if I understood correctly you just started checking it out?), there's probably something that stands out.

also, meaning .mozilla/seamonkey - as .mozilla could also have e.g. firefox (or did firefox change their directory?)

yes yes

so now i freed 2,2GB from my ~/.mozilla/seamonkey/ and my ~/.cache/mozilla/seamonkey/ in total

that 2,2 was cache plus what else?

thus actually i could run my savefilesize at 1,5GB with comfort

Seems that gcc 11.2.0 does not use CMOV instructions at all

no matter what CPU target

njsg, ~/.mozilla/seamonkey/ and my ~/.cache/mozilla/seamonkey/

also I cannot get gcc 14.2 to produce CMOVs when building my spreadsheet program

todays time is spare but during the next days i will look at seamonkey on another distro and if there is somthing like that either i let you know..

tomman: Maybe the code generator does not produce CMOVs at all? Maybe some libraries have inline assembly that erroneously contain CMOV instructions for i586 target?

Sompi: In my case I was crosscompiling with mingw

btw i was testing it here on an alpha release but from the current slack-repos so no wonder that something can go wrong but maybe i can find out why..

...which on Debian is still GCC 12

but this i will discuss at the porteus-forum

sorry though njsg fif i made you some hedaches ...

and seems that gcc 14.2 refuses to compile anything that has a call to gets()

I have a use case where it is absolutely not a problem

grobi: no worries, just trying to figure out where to look for. the figure sounded too large.

hmm, it still compiles when I use -std=c99

but again, it does not produce CMOVs

maybe the code generator itself does not use them at all

tomman: What libraries did you use with the problematic project? Are you sure that the project itself does not have inline assembly somewhere, with CMOV instructions?

not at all

And I built the libraries myself too

njsg, i agree !:)

it's a Super Mario 64 port built from this: github.com/sm64pc/sm64ex

and also built the dependencies, SDL 1.2, GLEW and libiconv

if I ask for a Pentium 4 executable, I get something with SSE2, as expected

if I ask for a Pentium III executable, I get something with SSE, as expected

if I ask for a i686, I get the CMOVs and nothing else

if I ask for i586/pentium... I get CMOVs and a executable that crashes and burn

even with -O3 I cannot get the compiler to output CMOVs

(Originally I used prebuilt deps from mingw repos, but that netted me a executable that would run on a Pentium III, unless a gamepad was plugged, then it would crash due to some Netburst-specific instruction getting involved, thus why I decided to crosscompile the deps too)

hmm, actually no

the -g flag seems to disable cmovs

when I remove the -g flag, it produces cmovs

but not when the target architecture is i586

only when it is i686

at least that's how GCC 11.2 works

and 14.2

so the bug is PROBABLY in some library, and not in the code generator

which C library?

I don't know

i686-w64-mingw32-gcc (GCC) 12-win32 is what I got here

(current Debian stable ships GCC 12 for everything)

hmmm, wonder what does that -g flag

It adds the debugging symbols

ah, that

And apparently also disables CMOVs

I'll try with Devuan's GCC 12.2 too

Hmmm, I should try a build too, but I will also check what additional CFLAGS are those makefiles adding

SDL's config.log already tells me -g is there

it refuses to compile to 32-bit, even with -S, when it does not have the 32-bit libraries...

CFLAGS = -g -O2

apparently libiconv defaults to those too

what is the apt package name for 32-bit cross-compiling libraries?

mingw-w64

actually it does not seem to have any libraries at all...

I'm talking about Debian and gcc

I remember that there used to be a package named build-essentials or something and it contained most development libraries, but it does not seem to exist anymore

yes, on Debian, mingw-w64 installs both 32 and 64-bit mingw crosscompilers

ah, that one is for native builds

I'm compiling for Linux

build-essential then

but it does not exist

build-essential:amd64/bookworm 12.9 uptodate

build-essential:i386 not installed

that's on this Debian stable box

so it was build-essential, not build-essentials!

I probably also need build-essential-i386

Not really, although the rare times I need a native 32-bit Linux build, I use a chroot instead

far less hassle that way

Hmmm, GLEW's makefile does default to -O2 only

no -g unless you're specifically aiming for a debug version

and sm64ex itself has no -g either (or -O2)

no, wait, it does

their makefile has like a zillion of BLAH_FLAGS that eventually get appended into the actual CFLAGS

I installed build-essentials and it still does not find any libraries...

and why does the SET command print out some weird source code instead of environment variables

debian has gone too far

...the hell, sm64ex has literally one set of OPT_FLAGS for _every_ source file

how do I see the environment variables

some have -O2, some have -O3, some have -g, and there is even CPU-specific stuff (for example if you're building a native N64 ROM)

it still refuses to compile...

Why does it even need those libraries with the -S switch? It's not going to link anything anyway

Hmmmmm... gcc has a flag to enable "conditional moves", -mcmove

...why

anyway, a red herring - nothing is using that flag anyway

normally it doesn't seem to use cmoves for 64-bit targets

only for 32-bit

And in any case, that option is not for x86 targets

only for some more exotic stuff like... SH-4

ok, now I finally got it to compile a 32-bit object file and generate its disassembly

With the default optimization settings it does not produce CMOVs at all. With -O3 it does, but only for -march=i686

With -march=586 it does not use CMOVs

gcc 12.2

tomman: SH-4 less exotic than or1k :D

Outside the Dreamcast, who even bothered with SH-4?

and gcc 12.2 seems to produce cmovs also for 64-bit

Sega only did because their CEO was best golf buddies with Hitachi's CEO

...OK, I have a digital TV tuner box with a SH-4 based SoC...

actually those other gcc versions produce cmovs for 64-bit too. I must have typoed something previously

this flag looks promising: -fif-conversion

and according to the manpage, there should be a -fno-if-conversion

that one can generate CMOVs, and seems to be architecture independent

it's enabled by default at pretty much every -O level but -Og

yarchive.net/comp/linux/cmov.html ha, Linus hates conditional moves :D

I cannot get gcc 12.2 to produce conditional moves even with the -fif-conversion flag

I will try later a build with that, need to get the ol' Socket 7 rig plugged into action

but first... some lunch~

I still think that the cmov instruction comes from some inline assembly in some library

i don't dare turning on my socket 7 rig :D

because the code generator clearly avoids that instruction when the target platform is not at least i686 or newer

I tolerate off topic and am happy to do it one or twice a day too but if you want to discuss ancient cpus and compiler instructions in detail pleaese dd it elsewhere. It's not even remotly useful for compiling SeaMonkey.

if i was up for community building this would be a great time to highlight that I am always on the forefront of off-topic conversations!

but i am gonna make a blog instead :P

hope everyone is doing well

lfs aquired

0511|19:07:02 < Sompi> how do I see the environment variables <-- like env, or the ones present when gcc is executed?

happy to go elsewhere for that too, although I don't have anything else to offer as a comment right now other than the above and that, at a last resort, I'd probably try to grep compiler tools directories for what that'd be in assembly.