Links to local documents (as opposed to targets inside local documents) are not checked

[martindholmes]

Commit a5f2f9e adds a test for this bug. We should not only be checking:

<ptr target="thing.xml#subthing"/>

but also plain old:

<ptr target="thing.xml"/>

We should be using doc-available(), presumably. The question is whether we limit this to documents with known XML extensions. Checking binary files is notoriously hard (http://stackoverflow.com/questions/2917655/how-do-i-check-for-the-existence-of-an-external-file-with-xsl) but we could also have a list of known text file extensions and use unparsed-text-available() for those.

[martindholmes]

Partial fix in commit 555508e. It now checks the existence of linked files using doc-available, but that will obviously open up the binary-files can of worms. I'll add something to the test to trigger this, and then work on a solution.

[joeytakeda]

MH and JT discussed this today re: binary files. We decided that we'll have to do this with a combination of ANT and XSLT. In another template, we create a text file that lists all binary files like so:

Full/Path/To/XML/Doc | path/to/binary | full/path/to/binary/file

Then we will have an ANT task (like ant-contrib?) that parses the text file and for each line uses properties set through RegEx mappers to check for the existence of the file (likely using the ANT available task), which would copy out the lines where the binary files are not available. Then, we run a second XSLT that transforms and injects the text results into the diagnostics output.

[martindholmes]

After some discussion, we think the only sensible approach here is:

1. XSLT creates a text file consisting of a list of binary docs which are linked. Each item in the list could have a pipe-separated tripartite structure (full path to target file, full path to source file, original link path).
2. In a subsequent ant task, ant uses a regex to parse off the first bit, and then uses its available task to check for existence. If it's not there, it copies the full line to a second file.
3. A second XSLT process re-works the diagnostics output page to add information from this second file to it.

[martindholmes]

I've taken the Release blocker label off this because I don't think it should be a blocker right now. We can go for an initial release without it.