with an ugly default to ~/.mozbuild . Previously, all the stuff goes to MOZ_OBJDIR (autocreated at "mach configure" time), now we need to define yet another dir. Why not default to @OBJDIR@/.mozbuild ?

that hasn't been the case for a while now

and its because backports of mozilla bugs

Thank that they at least warn that they are about to create a subdir in your home directory, and allow you to interrupt with Ctril-C. Excellent design! </sarcasm>

its the same thing Mozilla proper does

I know.

I used export MACH_USE_SYSTEM_PYTHON=1 in both versions 19 and 20 ( and I guess 21 too ) but build with "--with-system-icu" let the build process fail

norgo: have you tried to FIX system icu?

no I afraid I am to stupid dor this, I only commented this option and build process worked for me

buc: sent you messages

nsITobin Do you think Bug 1389738 can go in? Seems pretty obsolete per comment:

// NOTE: nsIHttpEventSink is only used for compatibility with pre-1.8

// versions

Just checking the tree a bit for missing up to 60 still which can be applied.

hold on sorry

i was doing dishes

frg:

nsITobin no rush

seems to indicate the same

cause this is where it is commonly used and documented to be a PROBLEM if not specified despite being stubbed out

so unless an adblocker is using it I say sure but keep it in the queue in case it needs reverted for a cycle or two cause its gonna be extension compat this will affect.

nsIChannelEventSink should be in use so yeah

I put it in and we will see if there is fallout in the prerelease. Suspect uBlock is not using it and nether is abp if you still use a fork of it.

do it, add a point on the relnotes, that nsIChannelEventSink should be used instead for any affected extensions which there almost certainly are none.

I doubt ABPrime works on SeaMonkey at this point considering it was fixed up pre-bootstrap abp with post-bootstrap patches and it wasn't entirely happy on UXP's xpcom feature level

xpcom interfaces feature level*

I need to prod the mangled thunderbird amov3 api see if i can get a scrape going

frg: check the rss feedhandling

for any usage

i don't know if it is now but this sparked some vauge impression of early firefox feed reader of which SeaMonkey's navigator feed reader was based

so dunno

frg: do you have a follow up patch for the toolkit search service as it at least at the state on xr uses nsIHttpEventSink

nsITbin checkxr and saw no references. Improt export tools 3.3.2 has a refernce but also tot he new one.

Oh the search service. I need to check this so postponed.

the feed reader is clean tho

in 1.1 (xpfe 1.8) sroaming made use of it.. then toolkit search made use of it.. rss never has.. and that's basically been the only real consumer in the tree of it since.. toolkit search service

nsITobin search service is covered in the bug

leaves the import export tools but I think easy fix there too and can just be removed

what license is import/export

onRedirect is not used in the code.

GPL

damn

if it was MPL i'd say let's just import the export tools

Claims to be compatible with TB 60 too so it should work without modifications. Wasn't sure if aIID.equals(Components.interfaces.nsIHttpEventSink) || throws but probably not then.

here's a test.. leave the interface but noop it see what happens

but no doesn't look critical and anything you'd use it for has a replacement interface

might be interesting to see how this turns out, cookies.sqlite get deleted on browser start, forum.palemoon.org/viewtopic.php?f=3&t=31907

DAMN IT

why can't I log into copr

and just get a 400 error invalid transition id

it's them

not seamonkey or anything else

therube: its just typical Moonchild apply Mozilla security patches that have no conception of being for a xul platform anymore and not testing.. This is shit I'd catch regularly when doing my quality assurance pass during rel-prep-week

Moonchild was a security expert.. PRE-PALE MOON and has done NO seucrity work outside mozsec patches for over a decade and every 3-4 months Moonchild would blindly apply a sec patch without testing and ship it.

this is just that

nothing to write home about

is the change that busted them

I have another theory

sabotage

because over a month ago I announced my Firefox Revenge plans for Pale Moon and Modern Firefox and every time for three years I have announced shit either I get attacked or the platform suddenly breaks with some untracked flaw.. after three years I am done considering it coincidence. Also means I can basically vaporware him into destroying his own platform so there's always that

that aside, still might be interesting to see how it turn out, just what the cause was

he says he doesn't understand why this breaks that .. my vauge intuition says maybe by killing it if it doesn't have a cell at all (whatever that means) is expected behavior for this era of code so returning early when the code isn't expecting it is causing either an aborted or fallback response like deleting the cookie store or something

mozilla TRIES to fail silently in later interations and UXP has a touch of that

isn't expected behavior*

one of Moonchild's biggest failings when working with Mozilla code is he doesn't GET that despite what the code says in isolation every change impacts the entire platform especially in key areas like spidermonkey or dom or layout or xpcom..

therube: assuming debug isn't likewise busted you could attempt a debug build of UXP if you wanted to find out but meh

that's beyond me. i'll leave it up to them & see how it plays out.

Nuck-TH tho if he stays eagle eyed and keeps pointing out MC's mistakes he won't be welcome there in the long term

yesterday I reported that seamonkey build fails if option --with-system-icu is used. Seems to be fixed now by 1692945-1-87a1.patch. Todays build of 2.53.21b1pre worked again with enabled option. For info only

norgo: love when that happens

don't you?

btw what makes someone support postgres AND sqlite but NOT mysql/mariadb

oh right, when you want to make sure the resulting downstream either jumps through enough hoops or has limited capability.. But what do I expect from redhat software.

I've no experiences with redhat, I'm using opensuse, slackware and archlinux

norgo: feel free to ignore me unless its important .. my rants are leaking today

norgo This was not a new patch. Maybe the newer system icu now needs nasm.

oh i bet that is the case.. yasm has fallen completely out of use outside mozilla-code largely

and yasm and nasm code aren't completely interchangeable

hello frank one week old or am I wrong?

norgo depends. Checked into gitlab for 2.53.20b1 a week ago. If you built only 2.53.19 and not used the patch queue then its one week old. Ripped out yasm some time ago for the .20b1 prerelease.

re: copr yeah their single signon is partly busted .. i did get in tho and then copr automagically became aware of my session.

frg: I build 21b1pre on 4th Dec and today with exactly the same system

then its sun spots. No changes here.

hey frg hg.mozilla.org/services

hidden

but still there

frg: Btw. wanted to say thank you for all you do (seamonkey) :-) very good job

norgo We try but it is becoming more and more of an uphill battle

it would be terrible if seamonkey die

yay, Imgur login is BROKEN on SeaMonkey

ReferenceError: currentStep is not defined

TypeError: form.submit is not a function

wonder what stupid new feature they deployed, but now I logged out and can't login again

90% its dynamic improts class fields or big int these days.

I can't even force-submit the form via JS because they somehow overrode the submit() method

^imports

...with a button named "submit"

hence the ".submit is not a function"

HOW DO THEY EVEN!?!?!?!?!?!?!

because xul outside xml doesn't work see: html5

document.getElementById("signin-form").submit()

i am cloning every HG repo that looks important especially services and webtools

this is completely broken, thanks to Imgur top meme engineers

and then I will generate bundles and put up on the xulorg

hell, even changing the stupid button to a ordinary submit doesn't work, since it eventually triggers some script that calls... form.submit

tomman: with customelements it may not even be dirived from HTMLButtonElement it may be a generic HTMLElement with totally custom logic

for some FUCKED reason all those html elements do NOT dirive from HTMLElement

HTMLElement is merely the generic element

it sthe one missing piece that would make customelements a complete xbl replacement

curiously enough the form here is using standard HTML, although there is a bunch of jQuery in the way too

I see no trace of Google WebComponents® crapola here so far..

document.forms has exactly one form, named signinForm, with a unusable submit method because it's now... a child item, in this case the broken button

...wait, so having a element named "submit" in your form (that is not a <input type=submit> button) will BREAK form.submit!??!!?!?!?!?!

yep, managed to login using that

var f = document.forms["signin-form"]

HTMLFormElement.prototype.submit.call(f)

the second one will call the real submit method on the form object referenced by f

...how does this dog vomit works on Chrome!? Accoridng to that other document it is supposed to be broken because... wat?

and THIS is why people end doing drugs, folks!

tomman: because Chrome does not ahear to any standard they didn't write or already had impled beforehand

and got the spec changed to match them

to the point where Google is replacing http with their own not-anything-like-http udp protocol

no html

no protocol

no browser

unless its theirs or approved by them

it is everything I am against webwise

The other problem with Memeur, er, Imgur is that occasionally you need to logout because otherwise you can't load the page with your uploaded images (instead you get a 404)

and that's what got me landing at this weird landmine today

tomman: fun

logout, login (or find out how to pick the lock, it seems), then suddenly the images page works again

frg: so the groups are hidden but their repos exist and I have cloned em but those repos removed from the top level are indeed gone forever

anything of value gone?

penelope

chatzilla domi but you have in-repo history from those..

cvs-trunk-mirror but i think i have a git conversion.. pyxpcom but i have a copy in gitform venkman is gone forever but has been busted forever.. I dunno where ldap was kept have to check the wayback machine for the url

repo for schema-validation and looks like a fuck ton of labs and misc projects

anything you didn't miss in the last years you probably miss never

perhaps

of course for my PRIMARY lines of research things like webtools in cvs with UMO and AMO v2 and the remora repo i have and the hg webtools stuff like mxr

dunno feels like more is missing but I may be remembering repos that were removed pre-pandemic

"Mozilla welcomes new executive team members"

just what we needed, more execs!

...wtf, Firefox has a SVP!?

...and a VP of "Firefox Product", whatever that means

> Ajit has years of product management experience from Square, Google, and most recently, Meta, where he was responsible for monetization of WhatsApp and overseeing Meta’s business messaging platform.

blog.mozilla.org/en/mozilla/internet-policy/google-remedies-browsers in the meanwhile, Mozilla sees its main money pipeline threatened and are... defending Google!?

we are living in very stupid times, as JWZ would say

All hail JWZ!

if i hadn't made so many mistakes in my own development I'd be trying to capture JWZ's attention

i figure he'd tell me to fuck off under most circumstances and perhaps rightly so