|
Notes |
|
|
|
I cannot reproduce this with the framebuffer frontend either 3.3 development tip or a 3.2 checkout.
Does this still occur for you in this site? or has the site changed? |
|
|
|
|
Hi,
I downloaded the latest CI build for Haiku and it does not crash, but I don't know if asserts are enabled there.
The crash still occurs with the release version of NetSurf 3.2 for Haiku so the site didn't change. I think already when I reported this we noticed it was reproductible only on Haiku. |
|
|
|
|
all CI builds have asserts enabled.
Closing this as I believe tlsa css updates fixed it since 3.2 release. If it reoccurs with 3.3 please feel free to reopen |
|
|
|
|
|
reopend as PulkoMandy says it is still happening on 3.3 |
|
|
|
|
|
Still happening here on commit fea40a0a78a70a29573bbdc940ea6f9ccdb6093b. |
|
|
|
|
|
I can't reproduce it either. It might help if you could reduce it to a minimal HTML/CSS test cases that causes the problem. |
|
|
|
|
|
|
|
|
|
Unlikely, there are no form elements on that page (or the page I get served, anyway), and the BeOS CSS rules are all for form elements. |
|
|
|
|
Ok, the problem is in the CSS for this page which sets the "html" element to be font:inherit.
This will reproduce the crash for me:
<!DOCTYPE html>
<html style="font:inherit">
<head>
<meta charset="utf-8" />
<title>Jupiter Descending</title>
</head>
<body>
</body>
</html> |
|
|
|
|
That test case works here.
HTML is the root element, so in LibCSS in css_select_style, it goes through the computed style and sets any property which is set to inherit to its initial value. For the font-size property, that happens in css__initial_font_size, which sets it to medium.
So I don't see how that assert can happen. Is your libcss up-to-date? |
|
|
|
|
|
Yes, I hit this problem with a fresh checkout and build of NetSurf, using the env.sh script. So everything is up to date (libcss is at commit f6f52757684709f7b09ed68b534226ae31633bce) |
|
|
|
|
|
Ok, after further investigation, this only happens if libcss is built with gcc2. I will be using gcc4 from now on because of duktape anyway, so we can probably close this. Unless mmu_man wants to insist on BeOS and gcc2 support, that is. |
|
|
|
|
|
Confirmed resolved in 3.4 release |
|
NetSurf-1695-debug-09-09-2014-07-18-51.report (29,354) 2014-09-09 08:29