This has got to be one of the coolest things I’ve ever seen about Mozilla as a platform.
This has got to be one of the coolest things I’ve ever seen about Mozilla as a platform.
Thank you, bug 375760:
And a big thank you to roc for writing that patch, and vlad for reviewing it too!
Woooooooooooooo Reflow Branch!
(Proof dbaron is awesome: I didn’t even notice for 6 days.)
I’m considering switching back to the 1.8 branch again. There’s a giant memory leak on trunk that it seems only I can see (yes, I’ve tried safe mode) that I’m having trouble narrowing down. Cairo is still incredibly slow on Linux. The tab scrolling thing that just landed is incredibly stupid, in my at least partly-qualified opinion, and interrupts my workflow a lot.
And to add to all that, ever since going to a music festival for 4 days and not using a computer at all, I’ve stopped really wanting to spend so much time on it in the first place, and I just want it to do what I want.
So yeah, I might switch back, to take some of the hassle away. That would also mean I’d file fewer bug reports…
I browse the web with
once, because I, you know, value my sanity. However, there is the occasional image that I do want to see animate. Does anyone know of an extension that lets me temporarily animate one particular image? If not, I think a context menu item (only visible on animated images, of course) would be a good solution for anyone that feels like making one.
I’m no longer running firefox trunk nightlies, having moved back to the 1.8 branch. It was nice having the new features, and I wish I could keep testing them, but the performance was just too bad. I was able to stand it for a long while, but then I had to test something on 1.8, and after that I just couldn’t go back. It’s amazing how a fast browser doesn’t seem all that much faster, but a slow one feels awful.
The perf bugs (mostly on Linux, from what I gather) have been sitting there untouched for a while, as people focused on the 1.8 branch, and went to XTech, and had vacations, and the like. Vlad says there will be work on them soon, though, within “the next few weeks” (on the 30th in mozilla.dev.tech.gfx). And indeed, one small fix went in yesterday, though not enough to make me switch back.
Hopefully “next few weeks” from the 30th turns out to mean starting at the end of this week and picking up next, so I don’t have to stay away too long.
Update: So it seems that that in fact, it wasn’t just psychological, there was a huge regression that happened on trunk while I was using branch. I’ve moved back to about half and half for now, because testing trunk is important and I know there aren’t enough people that do it, but I hope the cairo bugs get fixed soon.
A cool thing in Firefox that I figured out not too long ago:
Hope someone finds this useful.
I have one profile I use normally, one I use only for a few things (with special settings, userContent.css, etc.), one I use when I need to be logged in to the same site with two logins at once, and one I use for testing that I constantly delete and re-make to clean it.
To use them, I have a main button in the panel that just calls
path_to/firefox, and choose the profile I need, or if I need two at once I open a terminal and type
path_to/firefox -no-remote. This fails in a small way in one particular circumstance: If the partial update has already downloaded, opening a second instance will install it, which can lead to minor breakages (XML error in download manager, e.g.) in the first. Other than that, though, it works perfectly for me.
You know what would be a really nice extension for bug testing in Firefox? Something that hooks into the profile manager and lets you create a one-use profile without having to go through the four-step “delete old clean testing profile/create new/type in name/finish” sequence. When you’re looking for a regression range and opening with a new profile each time over and over, the savings would really add up.
So I’m reading the comments of a fastback memory usage post on Ben Goodger’s blog, and people are saying they rarely use the back button more than one or two times in a row, and storing 8 pages is maybe a little excessive.
And I just realized, I often use it not once or twice or eight times, but twenty or more. You see, I have a few html-based games I play, and when I’m moving around a map I can be loading over a page a second. And when I want to, say, go back and see the results of the last battle, that can be quite a few pages back even though it happened less than a minute ago.
Contradict that to slashdot where I open each story to read in a new tab and never use back at all, or google where, if I think I know what page I’m looking for but I’m wrong, I use it exactly once at a time. All in all, I think it’s pretty obvious that with such a wide range of usage even from just one person!, some sort of pseudo-intelligence is needed here.
As a disclaimer, I’ve never noticed any memory problems at all, though. Perhaps that’s because I have a gig of RAM and use Linux, and Linux’s memory management and multitasking/multi-userness generally pwns Windows’.
Argh, Slashdot got hold of the virus in the Korean contributed builds of Mozilla and Thunderbird. I was on IRC at the time they were discussing this, and I remember my first thought being “I hope slashdot doesn’t hear about this.”. And they didn’t even note the fact that those were contributed builds, not mozilla.org official ones, nor that ClamAV has now been installed, nor that the list of people with upload rights was already in the process of being reduced, to prevent just this kind of thing in the future.
*Sigh.* Time to go clarify the situation there…
Update: Slashdot seems to have gotten it into its collective head that the infected files weren’t hosted on mozilla.org at all. This is, of course, wrong, but if they’re thinking that, then I’m far too lazy to correct them.
Holy cow. I just came upstairs from reading, and I saw a dialog:
Deer Park has downloaded an important update, blah, blah, blah…. And I clicked Yes, or whatever it was to continue—just one click, mind you—and then the standard “You have more than one tab open, are you sure you want to close”, and then it started back up and I was using a newer nightly. The only thing it didn’t do was bring me back to the same pages I was on already. It’s amazing. I mean, I know that’s what it’s supposed to do, but wow.
I wonder what it would do though if I had installed it as root? I put nightlies in my home directory for convenience, but if I hadn’t, would it fail gracefully? Would it check to make sure it could do it first, before even downloading? That would be best… I should check on it.
I absolutely love the Web Developer Toolbar. One reason, that I ran into today, is that if I need to find out if a particular piece of (x)html is valid I only need to type
data:text/html,<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd"><html><head><title></title></head><body><p>sometext</p><img src='http://www.google.com/intl/en/images/logo.gif'/><p>moretext</p></body></html> and with two clicks of the mouse I can send it to the w3c validator. (The validator doesn’t have a form upload, so otherwise I’d have to write it, save it, and go through the upload steps on the site.) And since I have said data: url template bookmarked, it’s a total of 3 clicks and writing a minimal testcase for what I want to know–much easier than reading through the specs.
Note that the w3c validator does have bugs, so this shouldn’t be used for anything complicated or esoteric. Just the basics.
I’d like an extension that can uncompress various types of compressed files (zip, gz, tgz, etc.) and let me view the contents (assuming they’re viewable) in the browser. I ran into this today when I wanted to look at a wallpaper that someone had (stupidly) put in a zip file, butI’ve also realized it would be useful in bugzilla, for when people attach a log zipped.
I’d like it to work so it looks to me like the zip file is a folder, and it just showes me the file listing, from where I can choose what file to view. It would be nice if it could interpret relative uris inside itself, so, for example, webpage.zip/webpage.html could include webpage.zip/webpage.css transparently to me. Also, and quite importantly, it should have a big, easy-to-click button to let me save it if that’s what I really wanted, and it should not have to re-download it.
Here’s an idea. When someone finally gets around to implementing logout/etc. functionality in HTTP authentication, why not put the username in a little box at the right of the address bar, like where the lock is for secure sites. Not only would this be nice and usable, but it would also help stop phishing attacks of the form http://email@example.com/, since the username part, www.examplebank.com, would be moved away from the normal address and be visually distinct.
Bad Behavior has blocked 62 access attempts in the last 7 days.