Accept responses with no etag header

[indirect]

If the server does not provide an etag header, treat all responses as a
cache miss. This should allow 3rd party gem servers to more easily
provide compact_index results. /cc @mperham

See also sidekiq/sidekiq#4158.

[indirect]

@deivid-rodriguez have you seen this kind of failure on windows before? I'm confused about why it passes everywhere but windows 😬

[deivid-rodriguez]

Yes, I've seen those in the past in other places. I think it would happen randomly only on Windows when trying to rename a folder while being inside it.

[deivid-rodriguez]

I have no idea why the changes in this PR make the failure happen, but since the spec failing is the one being touched by this PR, it should be related to these changes.

[deivid-rodriguez]

I'll have a look at this tomorrow, see if I can figure it out.

[indirect]

Oh no, I think I figured it out. The existing test hard-codes /tmp/localpath, which doesn't exist in Windows. How did this ever pass on Windows before?? 😬 Hopefully this will fix it...

[duckinator]

This looks good to me. 👍

[fschwahn]

@indirect this patch is not in 2.2.0.rc.2 - with which version will this be released? This fixes a critical bug which makes it impossible to run bundle update for me (which incapacitates eg. dependabot). It's a bit hard to follow development now that the repositories are merged with rubygems (eg. no branches / no tags).

[deivid-rodriguez]

I forgot to backport this, sorry. Will set it for the next release.

[fschwahn]

awesome, thanks!

[deivid-rodriguez]

It's a bit hard to follow development now that the repositories are merged with rubygems (eg. no branches / no tags).

Stable branch for bundler-2.2 is that same as for rubygems-3.2 (the 3.2 branch), and tags are prefixed with bundler-. I know it's currently very weird, will get less weird as we improve things.

[deivid-rodriguez]

@indirect I wasn't sure of how to tag this PR. I went with bundler: minor improvement, but based on #3865 (comment) it could well be tagged as bundler: bug fix too? 🤔

[fschwahn]

Stable branch for bundler-2.2 is that same as for rubygems-3.2 (the 3.2 branch), and tags are prefixed with bundler-. I know it's currently very weird, will get less weird as we improve things.

Thanks, that helps. I didn't find the tags because github truncates the list, but searching works fine 👍

[indirect]

@deivid-rodriguez yeah, I would categorize this as a bugfix. It was the intended behavior, and this was fixing the implementation to match what we intended. 👍

[deivid-rodriguez]

Thanks @indirect!