[gs-devel] pdfwrite regressions & pdfmark problems

[Raph Levien]

Hi Peter,

   I wanted to follow up a bit on the pdfwrite regression issues. Jack
reports that many of the error cases are due to pdfmark failures when
running ps2pdf on pdf files. Coincidentally, we some pdfmark problems
in this week's support flow.

   In particular, in bug 567653, I found that NI (named image) was
unimplemented, and in such a way that it causes a typecheck downstream
(.PUTDICT expects a dictionary cos object, but it's uninitialized).

   I have a patch that works around this, but leaves orphan
(unreachable from Root) objects. Ultimately, I think we'll need to
implement NI, and also Namespace{Push,Pop}. I'm not sure what the
priority is, though. I suppose it depends on the customer. If we check
in my workaround, then hopefully a number of the error cases will go
away.

   Jack promises to get stdout/stderr into the regression output in
the next few days. In the meantime, the FAIL cases represent actual
rendering differences caused by recent checkins. Jack can nail down
the date of the offending checkin more precisely if that would be
helpful.

   Let me know if there's more I can do.

Raph