Tiny Tiny RSS: Community

Gogs (git.tt-rss.org/fox/tt-rss) download of tarball/zip of given commit blocked

While updating www/tt-rss FreeBSD’s port to use the latest commit (76dd74e0d9) it appears gogs/nginx is now blocking the ability to download a given tree’s tarball/zip. This might have started when git/tt-rss moved to fox/tt-rss. Last update to the port was 2019-10-09 commit 81bf1125a.

Trying to download 76dd74e0d9.tar.gz responds with a 401 auth. required error. It appears only master.tar.gz is allowed to be downloaded. Unfortunately FreeBSD ports framework needs a static download URL due to checksum validations. If master.tar.gz changes then the recorded checksum would fail.

Would it be possible to reconfigure gogs/nginx to allow this type of download or at least fox/tt-rss user/repo?

In the future, please at least search or skim through recent threads. Here you go:

Fair enough. Perfect. Thanks for the update.
I’ll deal with this accordingly.

this wasn’t actually intended, i guess the regexp for that URL matches (removed) numeric tags. i’ve removed the 401 altogether, with YY.mm tags gone it doesn’t make a lot of sense to keep it.

also, i’ve checked and freebsd port seems updated regularly enough from upstream code, thanks for that.

Excellent, I do my best to update it monthly.

very nice.

this change should make version display something more useful than UNKNOWN on packaged installs - https://community.tt-rss.org/t/versioning-changes-for-trunk/2974/27?u=fox - without hacks.

We hacked around it by changing version.php’s VERSION_STATIC’s value to the commit_hash.

I’ll revert that since it’s gone and start using version_static.txt with the same value.

Also, I think removing VERSION_STATIC might break your tests/ApiTest.php testVersion()?

yeah, phpstorm refactoring tool told me about it. those tests are more or less dead anyway, i probably should remove them altogether.

I think we might want to trim(file_get_contents(…)) or at least remove any trailing returns? Since the content is used in user-agent this appears to cause some oddities with http requests. redirects aren’t being followed, bad requests, http2 connections are cut short.

yeah, of course, trim() there would make a lot of sense.