I think nisheeth posted a basic doc on mozilla.org somewhere, but a
simple search couldn't find it. Anybody got a link?
fantasai wrote:
> I need to manual review some failed regression tests,
> and I have some questions. I can't find anything about
> them on mozilla.org--I only found out how to run them
> because Marc Attinasi posted about some new ones
> on n.p.m.style. So if there are some instructions I
> missed, please refer me to them.
>
> 1. What do the layout regression tests compare?
The layout regression tests compare the FRAME model for a set of
documents. Basically,
1. you need to build mozilla with debug on,
2. go to mozilla/layout/html/tests, figure out which tests your changes
might effect (block and table are the most mature, and most likely to
match.)
3. cd into those directories and type |rtest baseline|
this will create a number of data files that represent the frame model.
you will compare your changes against these files
4. make your changes to the code
5. run |rtest verify| this spits out a representation of the frame
model again, and compares it with the previous one for that document.
It tests resolved style, frame tree parentage, frame size and state.
6. any discrepancies need to be looked at manually. If the pages look
the same, it's likely not a problem. Also, your changes may themselves
introduce differences that are improvements, not bugs.
note: data files are not portable across platforms or machines, because
the results will depend on things like what fonts you have loaded. So
if you change your fonts, screen resolution, etc, you must re-build your
baseline. Also, you cannot change the window size while the tests are
running, because you will alter the size of the root frame.
Last I checked, the tests ran fine on linux and windows. I don't think
anybody has built a script file for mac yet.
Hope this helps -
Steve
>
>
> 2. If I were to manually go over some regression tests,
> would I just load the pages in Viewer? A lot of them
> aren't complete HTML documents, so I was wondering.
>
> 3. Also.. do I need a debug build to run the tests?
