Meeting minutes 2012 03 27

Attendance: Robert, Jason, Jarno

Item 1: 1.6 Milestone 3

Released last week.

Item 2: 1.6 Milestone 4 plans

Robert / Jason: Work on testing

Jarno: We've started getting new RFE's. Will take a look at those.

Item 3: Test plan

Jarno has a bunch of automated tests running. Discuss post-processors to clean up IDs to make comparisons easier; clear that it works for single-file output, unsure if chunking rewrites HTML and makes use of generated IDs. But seems it should not.

When looking at tests: have unit tests for a specific feature / function - if it doesn't work, match isn't one-to-one, feature is broken. Then have generated stuff like the IDs - need to check if the link still works; rewriting the ID is a way of verifying that the link still works. Last have changes that are not a problem - can change something in FO that looks much different but where PDF is exactly the same. Such as extra redundant block elements or a different way to specify the same font size. Could rewrite the FO to remove extra blocks.

In a lot of cases - old test case was full of wasted resources. Converts everything to everything. Could have just run pre-process and checked that. All coderef is in pre-process; if that breaks, every transtype breaks, otherwise it's the same as any other codeblock for the output. Would be good to clean this up. In a lot of ways ... preprocess has the most code and the most places where things could break. Best to focus there. That's where unit tests come in. JUnit, etc. Could also see about XSLT Unit tests.

Have test data up in Travis - free hosting for OS stuff, don't have to use it but it's an option. Can also test with other tools there. Don't need to do continuous integration; developers can run the tests locally.

Would like to focus on ways that we can speed up all of this up - spend less time testing things that just waste resources. If something is broken in preprocess, it is broken in output; do not need to compare output for the preprocess unit tests.

Currently - Jarno's automated output goes into a parser, XML Unit Test compares the parsed old output to the parsed new output. So, attribute order / etc is irrelevant - just comparing for significant XML differences. Used for preprocess output and for XHTML output.

Item 4: Milestone releases

Has been suggested that we release less often. Jarno likes the 3 week cycle - like to get things out. But if we can get nightly builds this might get rid of the concern - somebody can get the nightly build (clearly less tested), or the less frequent iteration test build.

Item 5: PDF questions / docs

Many questions coming to dita-ot-user list about PDF, without much response. Uncertain how to follow up on that. Problem today: to extend PDF, you really have to read the code, then write code. Jarno working on a PDF style generator that is getting better. As soon as get that updated again, may be able to get some of that done more easily - less advanced users could just generate a style.