[Groff] Re: groff regression tests

[Gaius Mulley]

Werner LEMBERG <address@hidden> writes:

> PS: What about your grohtml changes? :-)

Hi Werner,

well good news and bad :-) on the good side what has been done:

   *  retrofitted my changes into 1.19.2 cvs version.
   *  these include diversion handling for grohtml (remembering the
      troff state).
   *  -mwww can also handle style sheets and webpage.ms now exploits
      this. Grohtml can now handle the title in the webpage!
   *  also created a dejagnu infrastructure (heavily borrowed from GCC)
      for groff and I'm slowly populating some regression tests
   *  fixed a couple of minor bugs

on the bad side:

   *  I haven't finished yet!
   *  groff could use some minor libdevice changes to handle the
      proposed expected vertical/horizontal position glyph detection.
      This really shouldn't take too long and is quite easy coding..
      it could also be further improved with a mechanism to assert the troff
      state.
   *  I've found 2..3 bugs in grohtml which cause ugly output. These
      bugs are quite visible. References sometimes break (.IP), author and
      institution generate unnecessary line breaks in AU, AB when using
      -ms and .IP sometimes breaks when using point size > 10.
      And tables generate unnecessary <p> </p> pairs - as spotted by
      Larry.

my preference would be to push forward with making the regression
tests stress test the troff ditroff output destined for grohtml.
This way we can all populate the regression tests with our own nemesis
code and share tales of woe :-). It will also serve as a really good
method to quickly check that any change has not broken some obscure
usage. What do you think? How desperate are you to release the next
groff version?

Gaius