Undocumented <& sequence

[joelspadin]

I am using Mako to generate some Devicetree code, and I was surprised to find that

property = <&node>;

gets turned into

property = &node>;

when run through Mako.

I cannot find anything on https://docs.makotemplates.org/en/latest/ that documents <& (less than, ampersand) as a special sequence, but I do see it in a couple example files in this repo. Is this a feature that should be documented?

[sqla-tester]

Mike Bayer has proposed a fix for this issue in the main branch:

dont consume ampersands https://gerrit.sqlalchemy.org/c/sqlalchemy/mako/+/5561

[zzzeek]

from all indications this appears to be an erroneous character that was committed to mako 17 years ago as part of its original construction. I can find no use for this ampersand character. seems a little strange

[joelspadin]

Weird. Thanks for looking into it!

[zzzeek]

looks like it's from myghty since as you mention it's in the example files, it's in the myt files

[zzzeek]

ive released 1.3.6, however if this causes any regressions we might have to revert it, as that's been a thing lately

[joelspadin]

Thanks!

[git-bruh]

Hey, this seems to break firefox builds as a lot of their autogenerated code seems to rely on this quirk: https://github.com/mozilla/gecko-dev/blob/3b1f061102e8137e49c76382170045d9451e4406/servo/components/style/properties/properties.mako.rs#L8

This isn't a problem for now as firefox's build system use an older, bundled mako version. I faced this as I was forcing it to use my system's mako package which is newer than the bundled version.

[zzzeek]

firefox browser uses mako in their build system? unbelievable

anyway firefox should fix their templates when they upgrade, this was a real bug