Breaking change(?) to algorithms

[djn-ams]

At some point, something changed with how algorithms (specifically, captions, I think) are generated. The following compiles fine with TeX, but causes errors with TeXML:

\documentclass{amsart}

\usepackage{algorithm2e}

\begin{document}

\begin{algorithm}
  foo
  \caption{bar}
\end{algorithm}

\end{document}

Specifically, it will cause the following errors (main.tex being the name of the test file):

! Missing \endcsname inserted.
main.aux, l. 2>
                \@writefile{\ext@algocf }{\contentsline {algocf}{\numberline
{}{1}{\TeXML@caption }{ltxid1}}{\thepage }}

! Missing \endcsname inserted.
main.aux, l. 2>
                \@writefile{\ext@algocf }{\contentsline {algocf}{\numberline
{}{1}{\TeXML@caption }{ltxid1}}{\thepage }}

! Missing number, treated as zero.
main.aux, l. 2>
                \@writefile{\ext@algocf }{\contentsline {algocf}{\numberline
{}{1}{\TeXML@caption }{ltxid1}}{\thepage }}

! Missing { inserted.
main.aux, l. 2>
                \@writefile{\ext@algocf }{\contentsline {algocf}{\numberline
{}{1}{\TeXML@caption }{ltxid1}}{\thepage }}

))
! Emergency stop
! You ran out of file, mate!.
<unknown>, l. 1>
                 \@writefile{\ext@algocf }{\contentsline {algocf}{\numberline
{}{1}{\TeXML@caption }{ltxid1}}{\thepage }}

You ran out of file, mate!

Note that adding the algorithm package back in fixes things, i.e.,

\documentclass{amsart}

\usepackage{algorithm2e}
\usepackage{algorithm}

\begin{document}

\begin{algorithm}
  foo
  \caption{bar}
\end{algorithm}

\end{document}

will compile with TeXML. (Also note that this breaks compilation with TeX, so you have to use a nasty hack to get both TeX and TeXML to compile)

A cursory look shows that we've generated and posted/published papers before using only the algorithm2e so it seems like some change has broken the behavior recently. I don't see any recent changes to the algorithm package shims themselves, so I suspect it's somewhere else.

---

Whoopsie, I made a minor mistake. TeXML works fine if algorithm is loaded after algorithm2e. It still breaks if algorithm is loaded before algorithm2e.

---

Papers affected

• https://github.com/AmerMathSoc/mcom/issues/1012
• https://github.com/AmerMathSoc/mcom/issues/1022

[djn-ams]

Addendum: I made a big mistake. Loading algorithm after algorithm2e breaks many of the commands. I do not advise doing this.