Prevent file system paths from being included in mml3 extension.

[dpvc]

Webpack will replaced import.meta.url by a literal string for the path of the file on the system where it is being run. That means files that include import.meta.url will have references to the host file system where webpack is run (in our case, the travis-ci servers). The beta.3 version worked around this issue in the ts/components/loader.ts file by replacing #root/root.js with one that uses / rather than import.meta.url; but ts/input/mathml/mml3 also includes references to import.meta.url, and so the mml3 extension was still incorrectly including the full path name.

This PR fixes that problem by having the mml3 extension call on #root/root.js to get its root directory and making a path relative to that, rather than trying to look up the equivalent to __dirname using import.meta.url. That means, in particular, that we don't need the #mml3 pseudo-module only longer, so that is removed from the package.json and tsconfig files that included it.

In addition, we modify the root-pack.js file (that replaces #root/root.js when webpack is used) to look for a MathJax configuration MathJax.config.loader.paths.mathjax or MathJax.config.__dirname (the latter being set up by node-main) in order to try harder to find a good value for the root directory when used from webpacked files. Also, the esRoot() function was renamed mjxRoot(), as that is more meaningful.

Finally, the mml3.serf.json file also included full file pathnames based on my own file hierarchy, so the package scripts used to create it have been modified to end up using /tmp rather than my full path to MathJax-src, so at least that is more generic.

[zorkow]

I did not get it to work in the lab. Is there anything I need to do?

[dpvc]

I did not get it to work in the lab. Is there anything I need to do?

Hmmm. None of this should affect the usage in the lab (it is all for node-based use). It works in the lab for me, at least for this simple test case:

<math><mstack><mn>123</mn></mstack></math>

Can you say more about what is going wrong for you?

[zorkow]

That was a local issue. I rebuilt everything and it now works in the lab.

[zorkow]

lgtm.