WebKit weekly report #51 - back to normal schedule

Hello there! I'm now back home after BeGeistert. As you may have noticed if you read the development mailing list, there is a general agreement from the development team for a beta 1 release "real soon now", and after that, it will be time for R1. I will be coordinator for these, which means I will spend a bit less time on WebKit to take care of some other tasks. But I'm not giving up on WebKit, which still needs a lot of changes to reach release-quality. I was a bit less active this week as I had to recover from the lack of sleep at BeGeistert. But it's all fine now and I should soon be back at full speed. So, the change list in WebKit is a bit shorter than usual:

• Updated to the latest release of WebKit as I do every week. No huge changes there but a lot of small fixes, improvements and cleanups as usual.
• Started working on the implementation of data blobs. There was a debugger() call there and finally some website (osnews.com) started making use of the feature, so I had to remove that and try to implement the feature.
• Removed a hack that contributes to making our WebKit port slower than it should. It improve things but is still not perfect.
• Avoid popping up an authentication dialog for unknown authentication types. This happened a lot in Google Docs/Drive which uses the WWW-Authenticate header for things other than HTTP Basic and Digest authentication (it uses them for OAuth). While still slow, Google Drive gets slightly more usable with this change
• Added "Version/8.0" to the user agent so Google Drive stops complaining that we are an outdated version of Safari.
• Add an alert asking what to do when an invalid SSL certificate is received. This is not complete yet but the most boring part of the work (getting the error from OpenSSL and forwarning it through WebCore and WebKit up to the web browser) is done. The remaining work on making the UI better is on the web browser side only.

Some unrelated work on HaikuPorts:

• Added recipes for u_boot_tools and mtools. It is now possible to build the ARM version of Haiku from Haiku, making it much more convenient for me to hack on it. I still need to merge mmlr's fixes into linaro-qemu so I can test the resulting haiku image.
• Added a recipe for lnlauncher. I use this alternative to Launchbox and like it a lot.

And the bulk of my work this week is on Haiku. I started the long work of reviewing the ticket list for R1 and trying to move as much of those as possible out of the milestone, either by closing them or putting them in the "unscheduled" milestone for the R2 team to decide what to do with them. In no particular order the following issues are fixed:

• The media add-ons are built for the secondary architecture, now. In case your app wants to instantiate them directly.
• Custom cursors work again on video drivers with hardware cursor support. This was an old patch from looncraz which I merged.
• There are new haikudepot package for psiconv (a dependency for Abiword), droid fonts, most (an alternative to less)
• It is possible to build Haiku on an armv7hl host (this was missing from the list of supported architectures)
• Fixed various problems detected by Coverity in the network preferences. It is also actually possible to set a static IP now.
• Some UI fixes to CodyCam, PatchBay, BootManager.
• Improvements to BSecureSocket and BCertificate, and integration with BHttpRequest and BUrlListener. This is what makes it possible to get the "unsecure certificate" alert in WebKit but it is now also available to other applications.
• Fixed a problem with translators using the layout system for their settings window when used in Wonderbrush and Screenshot.

So, not a very long list of commits, and not a lot of code written. Most of my work was in reviewing the tickets and deciding what to do with them, and "just fix it" was reserved to simple tickets I could get done fast.