Make `read_resource_file` param simpler and more idiomatic.
`<P: AsRef<Path>>` is also what `File::open` uses as a generic type for
the parameter.
<!-- Reviewable:start -->
---
This change is [<img src="https://reviewable.io/review_button.svg" height="35" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/servo/10662)
<!-- Reviewable:end -->
Fix handling of borders and padding for empty/stripped inline flows
This forces fragment generation for empty inline flows, if they have borders or padding. Fixes#10533 and #2001. Also includes fixes for other bugs that were uncovered by this change; see the individual commit messages for detailed explanations. r? @pcwalton
<!-- Reviewable:start -->
---
This change is [<img src="https://reviewable.io/review_button.svg" height="35" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/servo/10643)
<!-- Reviewable:end -->
Disable tinyfiledialogs on non-OS X and Linux platforms.
I haven't been able to fix the appveyor build yet, so this gives us back a usable Windows build.
<!-- Reviewable:start -->
---
This change is [<img src="https://reviewable.io/review_button.svg" height="35" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/servo/10650)
<!-- Reviewable:end -->
Empty fragments may need to be layed out to draw borders, padding/background,
and insertion points. (Fragments that consist of discardable whitespace and
control characters, on the other hand, can still be discarded.)
This ends up preserving some useless empty fragments. It's possible we could
avoid this by storing some sort of flag on "important" empty fragments, so we
can discard the rest.
This fixes two problems that could cause scanned text fragments to end up with
incorrect LAST_FRAGMENT_OF_ELEMENT or FIRST_FRAGMENT_OF_ELEMENT flags:
1. If a single unscanned fragment was split into multiple scanned fragments,
then all of them would inherit its flags. We need to clear these flags,
except for the first and last scanned fragment.
2. When an unscanned fragment generated zero scanned fragments, we correctly
called `meld_with_next_inline_fragment` to transfer LAST_FRAGMENT flags to
the preceding fragment, but we didn't do anything to transfer
FIRST_FRAGMENT flags to the following fragment. We can fix this by calling
`meld_with_prev_inline_fragment` on the following fragment.
Factor out a new `meld_with_prev_inline_fragment` method that mirrors the
existing `meld_with_next_inline_fragment`.
This also fixes a bug in `meld_with_next` that was already fixed in the
`meld_with_prev` by @notriddle in #10419. The bug is that it was traversing
the inline context nodes in the wrong order. It should start at the outermost
enclosing node, since the fragments might be at different nesting levels under
some common ancestor.
This simplifies some logic in this patch and going forward. Note that the old code stored the
derived_from array as rust idents, but then used the parameter (not rust-ident-ified) to access
DERIVED_LONGHANDS. No other consumers actually seem to use the values in derived_from, so we
change the representation to the more natural thing.
Rust upgrade to 2016-04-12
This currently spews out a ton of ld errors of the form
```
/home/manishearth/Mozilla/.cargo/git/checkouts/mozjs-06d7f04b6dbb8a8e/master/mozjs/js/src/jit/JitFrameIterator.h:455: error: undefined reference to 'js::jit::SnapshotReader::readAllocation()'
```
cc @nox
<!-- Reviewable:start -->
---
This change is [<img src="https://reviewable.io/review_button.svg" height="35" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/servo/10585)
<!-- Reviewable:end -->
Remove unnecessary indirection in HTTP unit tests
Many of the HTTP unit tests use a custom request factory as well as a custom request that contained test logic. This is unnecessarily convoluted, and exists solely because the complete set of headers was unavailable until the request body was sent. These patches restructure the header manipulations so that the headers are available when the request object is created, allowing the test logic to move into the test factories, and enabling the deletion of almost all of the test request types.
<!-- Reviewable:start -->
---
This change is [<img src="https://reviewable.io/review_button.svg" height="35" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/servo/10555)
<!-- Reviewable:end -->
401 authorization UI then restart request/save successful auth creds
Step 7 of the NCSU student project Implement HTTP authorization UI
> make an authorization UI appear when a 401 HTTP response is received (StatusCode::Unauthorized) - in load in http_loader.rs, right before trying to process an HTTP redirection, use the new tinyfiledialogs library to make two prompts appear (username and password), then restart the request with the new authorization value present applied. If an authorization value was present and the response is successful, add the credentials to the authorization cache.
<!-- Reviewable:start -->
---
This change is [<img src="https://reviewable.io/review_button.svg" height="35" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/servo/10328)
<!-- Reviewable:end -->
Extracted Mako-based code generation invokation to separate .py files.
This breaks out some of the parts on #10586, that should be easily mergeable (hopefully pretty much a no-brainer really). The idea would be to let you review & merge it first, and then I'll complete the other PR rebase off of this stuff.
@bholley - I did like you suggested and broke it out for `geckolib` as well. The tests should also be running without problems (tested `./mach test-unit` locally).
<!-- Reviewable:start -->
---
This change is [<img src="https://reviewable.io/review_button.svg" height="35" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/servo/10617)
<!-- Reviewable:end -->
