Join GitHub today
GitHub is home to over 50 million developers working together to host and review code, manage projects, and build software together.
Sign updisplay list panic: 'index 26 out of range for slice of length 18' #10256
Comments
|
|
The destination port error is irrelevant, the important one is the slice index one which seems to be in some glyph code. cc @glennw |
|
Happens on master. |
|
I've seen this sometimes, but only on complex sites. This will be hard to solve without a reduced test case. |
|
This is probably a regression from some browser.html P1 optimizations I introduced. By the way, I investigated this before and concluded that the stack trace must be wrong as this doesn't appear in a function that actually performs bounds checks or calls any other function that does. Fun! |
|
I run into this bug fairly often, dump: https://gist.github.com/jrasanen/ad4d9b4a439bc9cb5e57063e5296c4b9 Opened servo with browserhtml enabled, pasted "https://v4-alpha.getbootstrap.com/examples/navbar-top-fixed/" into the URL bar and pressed enter -> immediate crash |
|
I wonder if the display list data for different iframes is getting mixed up. |
|
I'm having trouble reproducing this. I can't after trying each of the URLs here several times, with the latest browser.html. |
Tried with the very latest browerhtml & servo & webrender versions, still happening. STR: load firefox.com in bhtml, and click on links. Usually, after the first or second link, it crashes. I tried @jrasanen's page, but it doesn't crash here. @pcwalton I'll see if I can find the commit that regressed that. |
|
I can't reproduce on my macbook pro, only on my macbook 12. |
|
Thing's I have had in common when I get the crash, happened to me again while compiling a project: Higher CPU load, higher memory load (~600M swap used), I see at least one JS error in console, for Bootstrap it was "Bootstrap tooltips require Tether" |
|
Regression: 9fcf921 |
|
Yeah, I have a MacBook Pro, which explains why I'm not seeing it. :( This seems clearly nondeterministic… |
|
I get it regularly on MacBook Pro (Early 2015), 8 GB RAM, Intel Iris Graphics 6100. |
|
fwiw, still happening with the latest WR. |
|
@pcwalton, is there anything I can do to help you figure out what's going here? |
|
Attempt to load http://elm-lang.org reliably reproduces the issue for me. |
|
I also noticed that loading just that without browser.html does not seem to cause a crash. |
|
It's happening on my macbook pro as well. |
|
From a debug build, the key part of the stack trace:
|
|
Oh, it's obvious what's going on now. Working on a fix. |
avoid races. Addresses servo/servo#10256.
avoid races. Addresses servo/servo#10256.
Brand each payload message with the stacking context ID and epoch to avoid races. Addresses servo/servo#10256. r? @glennw
avoid races. Addresses servo/servo#10256.
Brand each payload message with the stacking context ID and epoch to avoid races. Addresses servo/servo#10256. Needs servo/webrender_traits#31. r? @glennw
Closes servo#10256.
servo: Update WebRender. Closes #10256. r? @glennw <!-- Reviewable:start --> --- This change is [<img src="https://reviewable.io/review_button.svg" height="35" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/servo/10462) <!-- Reviewable:end -->
STR: