520822 - iframe doesn't get redrawn when *reducing* main window's size

Status: RESOLVED FIXED

(Core :: IPC, defect)

Description

[Frederic Plourde (:fred23)]

When decreasing the main e10s window's width by dragging the right window border left, the inner iframe content doesn't get reflowed (see the attached video). However, increasing the window size does trigger effective reflow of the iframe.

Only tried on Win32 atm. Might not be the case on linux.

Comment 1

[Frederic Plourde (:fred23)]

Attachment: a link to the video

Comment 2

[Frederic Plourde (:fred23)]

I've tested on linux, and this seems to be a win32-specific issue.

Comment 3

[Frederic Plourde (:fred23)]

Attachment: patch v.1

This patch explicitly tells Windows to post WM_PAINT and WM_ERASEBKG messages to our child content window. It invalidates the whole window when its size is decreased.

bz: Tell me if you like that "#ifdef MOZ_IPC" at such a low widget level. Since everything worked just fine in the NON-IPC case, I felt like protecting this piece of code.

Comment 4

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 407028 [details] [diff] [review]
patch v.1

I'm not qualified to review this.  Rob, can you take a look?

Comment 5

[Rob Arnold [:robarnold]]

Comment on attachment 407028 [details] [diff] [review]
patch v.1

As I discussed with Boris on IRC last week, I'm probably not the best reviewer for this.

Comment 6

[Robert O'Callahan (:roc) (email my personal email if necessary)]

I don't understand why this is needed, and I'm afraid that it might just be covering up some other bug.

In particular, I don't understand why dispatching paint events would cause the window to resize. It might only be because we flush reflows before we paint, but forcing a paint would be the wrong thing to do if we really want to force a reflow flush.

Is the child content window not receiving WM_SIZE events?

Comment 7

[Benjamin Smedberg]

bz looked at this quickly on IRC: it turns out that because we don't have a nsIWebBrowserChrome for our nsWebBrowser, it always thinks the docshell is not visible and therefore delays and batches reflows on it... this will be fixed by bug 522533 which adds a nsIWebBrowserChrome.

Comment 8

[Benjamin Smedberg]

And so it is.