CI/tests: fonts SBL_Hbrw and AmiriQuran's addresses changed and now CI breaks

[Omikhleia]

At this date, the CI pipelines for tests currently fail:

/usr/bin/curl -fsSL https://www.sbl-site.org/Fonts/SBL_Hbrw.ttf -o .fonts/SBL_Hbrw.ttf
curl: (22) The requested URL returned error: 404

This has notably been seen in PR #2294 and #2278 - esp. for the latter, impending further work on it.

I think the "new "URL for this font is now: https://www.sbl-site.org/wp-content/themes/basket/fonts/bib-fonts/SBL_Hbrw.ttf

Do these folks have a proper/stable way to distribute that font? Can we question why SILE's CI and tests may be using such a "weird" font, while there might be more stable alternatives, distributed in a more stable wyay?

[Omikhleia]

FWIW, this font was introduced in 2019 via commit 399b31b

[Omikhleia]

Now the CI also seems to break for font AmiriQuran (URL in the makefile is to https://github.com/alif-type/amiri/ which has no 0.113 version, has only one version, much more recent, actually contributed by @khaledhosny if I read that well. It seems to be a fork of https://github.com/aliftype/amiri which has more versions... EDIT: Note the slightly different name, confusing...) -- Same questions as above:

• Do these folks have a proper/stable way to distribute that font
• Can we question why SILE's CI and tests may be using such a font, while there might be more stable alternatives (?)
• Why is SILE's makefile referring to a fork? (or what changed?)

[Omikhleia]

EDIT: Note the slightly different name, confusing...

And https://github.com/alif-type/amiri/ redirect to https://github.com/alicloud-sqa/amiri/ while https://github.com/aliftype/amiri still exists.
Very messy... And frightening...

[khaledhosny]

EDIT: Note the slightly different name, confusing...

And https://github.com/alif-type/amiri/ redirect to https://github.com/alicloud-sqa/amiri/ while https://github.com/aliftype/amiri still exists. Very messy... And frightening...

Very frightening indeed. I originally had the organization named alif-type, then I renamed it aliftype. Apparently github let someone create an alif-type organization again, and that person renamed it alicloud-sqa so alif-type now redirects to that.

[alerque]

Yes GH is terrible at handling renames safely. There is no mechanism for handling it well. Even if you were to squat on the old name yourself (the safest thing to do) you'd then break all the nice automatic redirects that happen when you do a rename.

The other situation people sometimes goof up is moving a repo from a personal namespace to an org. The redirects work great for that as long as that person does not ever fork the repo back into their user namespace ... in which case all the redirects are gone and old links stop going to the canonical new location.

In any event this is fixed for our CI and testing in 6da7bd0.