[LON-CAPA-admin] Problems with mimetex_valign on 2.8.0
Stuart Raeburn
raeburn at msu.edu
Tue Jan 20 03:46:45 EST 2009
Michael,
mimeTeX display of TeX expessions within <m></m> tags in LON-CAPA is
successful in Fedora 9/perl 5.10.0 on the i386 and x86_64
architectures running in my testing cluster. Commenting out the lines
you describe in lontexconvert.pm will indeed cause valign to be set to
zero, with the result that images generated by mimetex will not be
aligned with the text (which was the way things were pre-2.8.0).
This would seem to be an issue with the ppc64 architecture, which as
you say is not currently an officially supported architecture for
LON-CAPA. If you have a testing server available which has the same
ppc64 architecture as your production servers I could investigate what
the issue might be with socket creation for the HTTP HEAD request used
to retrieve the Vertical-Align value from the response header returned
by the initial request to mimetex.cgi. Alternatives to
IO::Socket::INET, e.g., using higher level packages such as LWP
introduce their own problems in cases where LON-CAPA's &xmlparse()
function is used (calls occur within Safe space), which is the reason
why IO::Socket is used for this.
Stuart Raeburn
MSU LON-CAPA group
Quoting loncapa_admin <loncapa_admin at johnabbott.qc.ca>:
> Hi,
>
> I've just been putting our set of servers through their paces before
> going into production, and I noticed that the Change Math Pref page
> was choking on the display of the mimetex box. Safari 3.2.1 would
> leave the iframe empty and report an error
> "http://kepler.johnabbott.qc.ca/res/adm/pages/math_example.tex?inhibitmenu=yes&texengine=mimetex -> lost network
> connection"
> while Firefox would try to download a 0-byte length file
> math_example.tex. A little further investigation found that this
> could cause big problems if the math display preference were set to
> mimetex or default.
>
> We're running something of an unusual configuration, Fedora 9
> running on ppc64 powermacs. The LON-CAPA specific packages have
> been built from source, while all the other packages are up-to-date.
> Since this is an unsupported architecture, I'm not sure if this is
> Fedora 9- or ppc64-specific, or just some weird quirk that happens
> on Mondays. I have been able to (I think) narrow down the problem.
>
> The mimetex.cgi binary is compiled correctly, and working (tested
> this by pointing browser to
> http://kepler.johnabbott.qc.ca/cgi-bin/mimetex.cgi?x^2+y^2, as well
> as calling mimetex.cgi from the command line).
>
> A bit of digging led to the following discovery: commenting out
> lines 249 through 279 of lontexconvert.pm (in effect forcing the
> variable $valign to be 0, bypassing the socket creation steps) gives
> a working mimetex (minus the vertical alignment, I presume).
>
> Has anyone else noticed this behaviour?
>
> Cheers,
>
>
> Michael Dugdale
> Department of Physics
> John Abbott College
> 21275 Lakeshore Road
> Ste. Anne de Bellevue, Québec
> H9X 3L9
>
> (514) 457-6610 X 5888
> loncapa_admin at johnabbott.qc.ca
> michael.dugdale at johnabbott.qc.ca
>
> _______________________________________________
> LON-CAPA-admin mailing list
> LON-CAPA-admin at mail.lon-capa.org
> http://mail.lon-capa.org/mailman/listinfo/lon-capa-admin
More information about the LON-CAPA-admin
mailing list