In Polygon, if you go to a contest, this link is almost unusable:
Specifically, I mean the "English (from working copies)". If you have a typical ICPC-style contest with 12 or so problems, it will almost always load for a really long time until I am eventually served with a CloudFlare "timeout" screen. (The one from packages works fine, but sometimes you want to see how the whole contest looks without committing your changes or packaging.)
My suggestion is, if there is truly that much work to do for the server, then run this PDF generation in some sort of background thread and show the user a progress bar.
This feature has never even worked for me. It always loads some weird error but preview in HTML works fine.
Even when creating an empty problem? There are some things that aren't supported for the PDF output because the resulting LaTeX is invalid in some way.
No, but I'm not sure when it doesnt crash. When I read the error logs it's like title too long (lol??) or some weird error with statement
Even if all the statements are working individually ??
It usually works with me if there is no latex problems in all the problems
Yes, I have this same issue as well
In my experience, you can open it in another page, do something else for 10 minutes, then go reload that other page. If you didn't save any statements in between, the document will be available almost instantly.
That said, I'd also like to see contest statements faster. On the other hand, I know LaTeX can get snaily.
I have observed that too. But the issue is most noticeable when you have errors that show up when you compile the PDF of the entire contest, but not when you look at PDFs of individual problems. In that case, you want fast feedback, not to do something else for 10 minutes.
I don't think it's LaTeX. I have previously compiled contest PDFs locally
to insert emojis in problem names, using exactly the package provided by Polygon. It only took about 10 seconds. Not exactly blazing fast, but also not the slowness we see in Polygon.Related: the real issue for me with LaTeX compile errors is the following.
The system tries more than one way to compile statements. Looks like one is latex + latex + dvipdfm[x]. And the other is pdflatex + pdflatex. Or something similar.
Anyway, without some care from the contest author, one succeeds only with eps images, the other only with png images. And perhaps both are needed for legacy contests.
My issue with this is: when way 2 is preferred and fails, the compile error comes from way 1, making the actual problem harder to find.