Bug 20121 – Cleanup of FrameLoader load() in a move towards sanity

• Component: WebCore Misc.
• Status: RESOLVED FIXED

[Daniel Jalkut]

I came upon the problem of trying to understand FrameLoader's variety of load() methods, and made an exercise of annotating/debugging my way to roughly understanding which methods are used, and under what circumstances.

I renamed many of methods in an effort to increase mental graspability, and hopefully as a setup for further cleanup iterations. I don't think that the vast number of overloaded "load()" methods was serving the intended value of such overloading.  Hopefully the new naming will give a better context for searching on how the methods are used, and for possibly folding the functionality further into more logical subroutines.

This patch also includes a couple smaller cleanups which are detailed in the changelog notes.

[Daniel Jalkut]

Created attachment 22398 [details]
Proposed changes to FrameLoader's load() methods, and changes to WebCore and WebKit clients of those methods.

[Daniel Jalkut]

I realized my comments above make it sounds like the changes are mainly in renaming, but there is also some degree of changing the signatures of the affected functions as well. In particular, where I spotted some opportunities to move towards a more clearly parameterized functionality.

[Daniel Jalkut]

Created attachment 22399 [details]
Updated patch to eliminate some stray tab characters, fix a typo.

[Matt Lilek]

I'm not a reviewer, but you use "nil" once in FrameLoader.h and again in the .cpp.  I'll also attach a patch that keeps windows building.

[Matt Lilek]

Created attachment 22402 [details]
Changes for windows

[Daniel Jalkut]

Thanks, Matt!

[Geoffrey Garen]

Comment on attachment 22399 [details]
Updated patch to eliminate some stray tab characters, fix a typo.

Just to give you some background, all these "load" methods came about when we merged the loader code, which had been sprinkled throughout WebCore and WebKit, into a single WebCore class. The merge illuminated a spaghetti-like structure (perhaps an image of the holy flying spaghetti monster itself?!) of similarly named and purposed functions. At the time, we named all such functions "load" as a way of saying, "We should probably refactor this code, removing and/or merging redundant noodley appendages, when we get a chance."

Anyway, I think your renames improve clarity, so I'll say "r+", but an even better solution would refactor the loader and eliminate some of these functions altogether.

[Geoffrey Garen]

Comment on attachment 22402 [details]
Changes for windows

r=me

[Daniel Jalkut]

Hi Geoff, appreciate the review and thanks for the clarification on the load methods. I can see how it's funny to have many of them change names again. But I do hope that it will help facilitate further refactoring. Maybe even by me!

[Mark Rowe (bdash)]

Landed in r35377.