Live data from Hacker News

Firefox 50.0

mozilla.org

21–30 of 205 posts

Re: Firefox 50.0

#21

Earlier quoted context omitted.

If you have a hard requirement to use a thoroughly obsolete GUI toolkit, you can for now fairly easily compile your own copy of Firefox for GTK2; it takes around 30-40 minutes on a laptop with a Core i5. (On Gentoo, there is a USE flag for it.)

If it is that bloody simple, why can't Mozilla be bothered to provide an official download?!

Possibly for the same reason they don't ship Tk compiled versions. Why should they target every GUI toolkit and complicate their build process which they then have to deal with supporting?

If it's that easy, you can probably find some third party that compiles and offers it. I'm not sure why you think Firefox needs to do it themselves.

Re: Firefox 50.0

#22
post #2

> View a page in Reader Mode by using Ctrl+Alt+R (command+alt+r on Mac) Took them bloody long enough. Now if only i was not stuck on ESR because GTK3...

> Now if only i was not stuck on ESR because GTK3... What's the issue with GTK3?

GTK3 itself has various regressions and is slower than GTK2 despite the architectural refactoring lending itself to a snappier experience.

https://bugzilla.gnome.org/show_bug.cgi?id=771708

Firefox GTK3 has regressions too, for instance the bookmark manager not remembering where you were last which works with Firefox ESR (GTK2).

https://bugzilla.mozilla.org/show_bug.cgi?id=1267863

Also GTK3's file dialog is subjectively a huge regression, but others may prefer it, just like Apple's finder changed over time. I still haven't figured out how dconf and gconf work with GTK3, not using GNOME3 as a desktop.

There are more issues with GTK3 but these are the most visible for those of us who don't use GTK3 regularly as part of GNOME3 and have been forced to use it via Firefox. The way I read the GTK3 bugreport it seems like the devs don't test GTK3 outside GNOME3 and hence do not consider it a priority. That one dev has been stressing that a compositor is needed and multiple answers by the reporter that they're using a compositor seem to be missed during reading on the other side. It's a weird exchange. I'm with the reporter. If GTK3 is not supposed to or not tested outside GNOME3, then this should be communicated so that everyone can make an informed decision to use something else or revive the Firefox Qt toolkit code.

Re: Firefox 50.0

#23
post #5
post #2

> View a page in Reader Mode by using Ctrl+Alt+R (command+alt+r on Mac) Took them bloody long enough. Now if only i was not stuck on ESR because GTK3...

If you want to avoid GTK3 and you're on Arch, there's a firefox-gtk2 [0] AUR package that compiles Firefox against GTK2 only. [0]: https://aur.archlinux.org/packages/firefox-gtk2

My local cairo-gtk2 build of Firefox 49 reliably crashes anytime you try to use the GTK file dialog. Does it work for you (arch or gentoo)?

Re: Firefox 50.0

#24
"Set a preference to have Ctrl+Tab cycle through tabs in recently used order "

Finally something that is an actually improved UI feature!

Not some "removed status bar and instead hover its info over text you want to read sometimes" or "moved refresh button to different place than before just to annoy you" or similar thing :)

Re: Firefox 50.0

#25
post #21

Earlier quoted context omitted.

If it is that bloody simple, why can't Mozilla be bothered to provide an official download?!

Possibly for the same reason they don't ship Tk compiled versions. Why should they target every GUI toolkit and complicate their build process which they then have to deal with supporting? If it's that easy, you can probably find some third party that compiles and offers it. I'm not sure why you think Firefox needs to do it themselves.

I think the third party situation is unlikely, unless they're shipping a fork like IceWeasel or stripping out all branding information. At the end of the compilation process in Gentoo it helpfully tells you (if you enabled the Firefox Branding) that you cannot legally redistribute a branded version of Firefox.

Re: Firefox 50.0

#26
post #21

Earlier quoted context omitted.

If it is that bloody simple, why can't Mozilla be bothered to provide an official download?!

Possibly for the same reason they don't ship Tk compiled versions. Why should they target every GUI toolkit and complicate their build process which they then have to deal with supporting? If it's that easy, you can probably find some third party that compiles and offers it. I'm not sure why you think Firefox needs to do it themselves.

For the same reason they haven't made 64-bit the default on Windows and provide two versions. There are many users for whom only one variant works reliably. If upstream, like in the case of cairo-gtk3, decides GTK2 is not supported anymore, it's a matter of chance whether your cairo-gtk2 build of Firefox works at all. For instance, for me 49.0.2 gtk2 local build on Arch crashes anytime I try to use the file dialog, while of course the file dialog of both GTK2 and GTK3 doesn't crash in other gtk apps on the same machine.

Somebody requested this and the reality has been ignored so far: https://bugzilla.mozilla.org/show_bug.cgi?id=1268234

Re: Firefox 50.0

#27
post #23
post #5

Earlier quoted context omitted.

If you want to avoid GTK3 and you're on Arch, there's a firefox-gtk2 [0] AUR package that compiles Firefox against GTK2 only. [0]: https://aur.archlinux.org/packages/firefox-gtk2

My local cairo-gtk2 build of Firefox 49 reliably crashes anytime you try to use the GTK file dialog. Does it work for you (arch or gentoo)?

After almost a decade of building Firefox from source I finally bit the bullet to switch to a precompiled binary last month. It's fixed all the issues that have started to crop up but leaves a bad taste in the mouth.

Re: Firefox 50.0

#29
post #23

Earlier quoted context omitted.

My local cairo-gtk2 build of Firefox 49 reliably crashes anytime you try to use the GTK file dialog. Does it work for you (arch or gentoo)?

After almost a decade of building Firefox from source I finally bit the bullet to switch to a precompiled binary last month. It's fixed all the issues that have started to crop up but leaves a bad taste in the mouth.

I had been using their binaries due to security updates and having to build at inopportune times for an hour or three (depending on machine), but like I wrote in a sibling comment, enabling the cairo-gtk2 backend resulted in crashes when trying to select a file with the gtk file dialog. So I bit the bullet and am struggling with the drawing regressions of GTK3.

Re: Firefox 50.0

#30

Also to note: Firefox can now upload directories. Granted, it's done by mimicking Webkit APIs, but that has the advantage that existing sites work without changes.

> Granted, it's done by mimicking Webkit APIs, That's ... not a bad thing. Chrome creating nonstandard APIs (IIRC this was for Drive?) on its own is a bad thing. Coming together and speccing ( https://wicg.github.io/directory-upload/proposal.html ) the API is a good thing. They seem to have specced more or less what Webkit had already implemented (plus some promise based stuff), but usually when a nonstandard API has…

> They seem to have specced more or less what Webkit had already implemented

The problem is that the WHATWG always just standardizes "whatever Chrome does".

This leads to the HTML living standard actually being just like Office Open XML, with one entity controlling all of it.

Post reply on HN