Eating tabs, wasting memory

[Kestutis Snieska]

Running Midori 0.5.7 ((null)) on Windows 2008 R2 Server

    Copying an open website URI to a new tab destroys the new tab ("eats it"), but the memory used for the creation of the tab is not released. After eating the tab, the tab from which the URI was copied is activated.
    The memory is not released even after 1 hour of not using the browser (tested for possible late memory release scheme -- if any).
     Does not depend on the protocol (could be http:, https:, about:, may be others).

    Steps to reproduce:
       1. Start Midori browser
       2. Open https://duckduckgo.com/?q=&t=midori (opens automatically if start page is set to about:search).
       2.1. check the memory usage of the browser
       3. Create a new tab
       3.1 check the memory usage of the browser
       4. Copy the link from the first tab to clipboard
       5. Paste the link from clipboard to a new tab
       6. Result -- the window displays only the first tab -- new tab disappears.
       6.1 check the memory usage of the browser
       7. if wanted -- create more empty tabs and then enter the same URI to it -- all will one by one disappear.

The steps 4. and 5. could be done by hand -- what is important here, that the URI would mach exactly.

      Happy resolving.

about:version

Midori 0.5.7 ((null))
GTK+ 3.6.4 (3.6.4) Glib 2.34.3 (2.34.3)
WebKitGTK+ 1.10.2 (1.10.2) libSoup 2.40.3
cairo 1.12.10 (1.12.10) libnotify No
gcr No granite No
Platform Windows NT 6.1
Identification Mozilla/5.0 (Windows NT 6.1) AppleWebKit/537.6 (KHTML, like Gecko) Chrome/18.0.1025.133 Safari/537.6 Midori/0.5
Video Formats H264 [x] Ogg Theora [x] WebM [x]

Netscape Plugins:

Google Update Google Update
Java Deployment Toolkit 7.0.510.13 NPRuntime Script Plug-in Library for Java(TM) Deploy
Adobe Acrobat Adobe PDF Plug-In For Firefox and Netscape 10.1.9
Java(TM) Platform SE 7 U51 Next Generation Java Plug-in 10.51.2 for Mozilla browsers
Silverlight Plug-In 5.1.30214.0

[gue5t gue5t (gue5t)]

The "eating the tab" part is actually expected behavior, as part of the ability to use the location bar to search through open tabs. Maybe the logic should be separated so that searching open tabs only switches to them if the completion entry to do so is chosen rather than if the url is identical. In the meantime you can use "duplicate" on the open tab (right-click it and select "duplicate tab") to get multiple open views of the page.

The memory usage is a known problem that we're meaning to investigate thoroughly. It may be the case that reference-counting on web views is not correct, or it may be that webkit is leaking the memory despite Midori doing what it should.