Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
[rawtex] mmaauth.cls #985
changed the title from
LaTeXML hangs forever (100% cpu usage)
May 3, 2018
Thanks for the report! I've renamed the issue (and reproduced the infinite loop with
I am also impressed you found the sty directory of arxmliv!
As a clarification, I am labeling this issue as an "enhancement" as we are currently not claiming that latexml can natively interpret all class and style packages that native tex can. But that is certainly the goal, and this report is quite helpful, especially given the minimal example!
Will take a more detailed look soon.
A quick remark, given that the loop is in
Certainly important to avoid the infinite loop though, so that needs a patch just for the principle of it.
Right, just noticed that you are affiliated with this project. However, to my regret, that was not me. Even when I know where it is I cannot build a google search request to find it. That was my team leader who found this archive.
Anyway, thanks for the instant response, hope the test case I've provided is helpful in any way. Cheers!
Yes, the page you found contains work done just about 10 years ago, and has been largely discontinued since - everything of value was brought over to the main LaTeXML suite, exactly here
That's why I mentioned that I was impressed you found that page. I am still periodically running latexml over the entire arXiv corpus, but any tangible results of that appear here as github issues / pull requests right away.
I took a closer look at the raw package interpretation prior to Bruce's upgrades.
The cause of the loop is subtle, as the loop itself is quite intricate and relies on a mixture of expansions and executions.
I could avoid the loop by adding the native definitions for
Given that, to my current understanding, LaTeXML's font machinery is on a rather higher level than latex's font internals, it does make sense to go the bounding route, as Bruce did, and avoid these concerns.
This does pose an open question about interpreting natively