Re: [Groff] Re: MSVC Port--Issue with groff Hanging

[Keith Marshall]

On Thursday 04 December 2003 1:11 pm, Jeff Conrad wrote:
> Keith Marshall wrote:
> > BTW, in the pic.ms case I cited yesterday, I also get normal completion
> > if I include a page number greater than the document length in the output
> > page range list;  I simply wasn't using a big enough number (>36) when I
> > was experimenting with it.
>
> At least we're getting about the same results. Not that it solves anything,
> but I guess it's better than getting different results for no apparent
> reason. Hopefully a fix in one port will work for both.

Hopefully.  We certainly do seem to be seeing similar results, except that I 
am seeing the "output error" message in every case, after interrupting a hung 
groff.

Although it is rather a contrived test, and there is probably no valid reason 
for specifying an page range for html output, (unless Gaius, or anyone else, 
can suggest one), I have also found that I can get the html example to 
achieve normal completion, without suffering the "broken pipe" condition, 
provided I include *both* page 1, *and* at least one other page with a fairly 
large page number; (-o1,10 seems to work for webpage.ms, while -o1,9 still 
results in the "broken pipe" failure).  However, if I omit page 1 and specify 
a range extending to a sufficiently large page number, then the plot thickens 
further;  instead of the "broken pipe" failure, I now get a memory access 
violation crash, (equivalent to SIGSEGV on UNIX).

I am already investigating a SIGPIPE issue (on Linux) with pre-grohtml, 
(although progress is slower than I might like, due to other demands on my 
time).  Perhaps I will also discover an explanation for this strange observed 
behaviour, but I really haven't a clue why the specification of a page range 
should affect the process in this way.

Best regards,
Keith.