[Gs-devel] Report on SourceForge bugs week of Apr 16 - Apr 23

[Raph Levien]

On the suggestion of Ralph Giles, I'm now sending these weekly
SourceForge bug reviews to gs-devel instead of the internal Artifex
mail alias.

416830: Auto-rotation was not being activated. Alex has a fix in code
review, just approved (with suggestions).

416973: We talked about this last week. The followup to the user is
good for the time being. In the longer term, I am not comfortable with
the ridiculously slow mode remaining in Ghostscript. Russell will
probably need to be involved to help figure out how to remove the slow
case. One thought that has occurred to me since last week is that, on
Unix at least, stdin might be set to nonblocking. Then, the "block at
a time" mode should be safe even when another application is piping in
commands and expecting responses on the stdout pipe. Implementing
nonblocking reads will require just a bit more code, including
handling the EWOULDBLOCK case, and making sure that the read handles
the case where the buffer is not entirely filled. The one thing I'm
less sure about is how to handle this reliably on non-Unix platforms.

417127: Looks like another stream problem. Igor has a patch in code
review, but I'm _extremely_ nervous about fiddling with the internals
of sreadbuf(). The potential for the whole house of cards to come
tumbling down is too great. I am reminded that documenting the stream
invariants still needs to be done.

417180, 417181, 417182: These look like duplicates of 413747. No
followup was posted.

417843: The user is reporting two bugs under the same number. The
first is that some Type1 fonts were being rasterized (in 6.50). This
is, of course, fixed in 7.00, which Alex pointed out. The second issue
is some type of cropping problem. No followup yet on the second bug.

418010: It may be that one of our web pages misspells "sourceforge" in
the description of how to get CVS access. I followed up asking for a
URL so we can fix it.

418241: The user complains that the epswrite driver always sets the
bbox lower left corner to (0, 0). This may be a side effect of some of
the bbox changes made to make sure that negative coordinates would
work. No followup.

418242: Yet another pdfopt failure. Alex followed up asking for a
file, but apparently it's not forthcoming. It's entirely possible that
this user's problem is fixed by one of the other pdfopt bug fixes.
Since the user is unwilling to provide a test file, I am comfortable
closing this bug as "invalid", with a suggestion that the user
replicate the bug with a non-confidential test file if it still
doesn't work in 7.01, or that (even better) the user is notified when
the pdfopt fixes get checked in to CVS.

418391: Text is "scrambled" in 6.50 and 7.00; 6.01 worked. Looking at
the PDF file posted, it looks like the text is (a) being rendered into
a bitmap, which is bad, and (b) positioned somewhat randomly on the X
dimension, which is also of course not good. I posted a followup
asking for the original PS file.

Note that of 10 bugs submitted by users outside Artifex, six did not
get followups. I'm obviously not happy about this. We need to be
responsive to reports as they come in.

Raph