integration test fails on complex forms updating multiple models

Oh, there was some outcommented debug code in my diff, here is a more clean diff

Sorry.

I have now made a patch according to "Contributing to Rails" article https://rails.lighthouseapp.com/...

Here it is.

Seems like current edge requires newer rubygems than I have on my Intrepid. I'll take a look at it when I upgrade my dev-box to Jaunty. I can at least see that edge does no longer contain the failing code, it has probably moved to the rack project.

However, looking at rack edge source code, it seems that Utils::Multipart.build_multipart in rack/lib/rack/utils.rb contains the essential fix.

In case someone is interested in a patch for Rails 2.3.3, here it is.

Jarl

I wonder why no one applies this patch to the 2.3.x branch.

I have to do this manually every time I upgrade rails (now to 2.3.4)

This is unpractical.

Jarl

Joshua, I doubt that this is fixed in Edge (as of May 2009). SInce the problem still exists in 2.3.5. Could you please have a look at it again. Otherwise please guide as to how I should show a failing unit test in edge.
In what file shoud I add a test? Or in what directory should I add a file?
then what (rake) command should I run to see test results.

I here attach a patch for 2.3.5

Jarl

I have now made a patch (according to contributor guide: https://rails.lighthouseapp.com/projects/8994/sending-patches). The patch is against the 2-3-stable branch.

The patch includes two unit tests that demonstrates the problem plus the fix of course.

Please apply this patch against 2.3 branch, will you?

Jarl

The above patch did not work for us. I've modified it a bit and it causes the app tests that were previously failing to pass. This patch is based on Jarl's. It applies cleanly to 2-3-stable.

Good to see there is some interest in this ticket.

Just of curiousity: Which "above patch" did not work? I have supplied several patches to this ticket. You say that your patch is based on mine (but not which one), and at the same time "the above patch did not work" while all of my patches are above your comment :-)

Another question: which branch is you patch against?

Note further that this ticket is marked "resolved" because Joshua believes this is fixed in edge. However it has not been verified by anyone yet, and even if it was solved I would love to see tests (in edge) that ensured not introducing the bug again.

Jarl

José, Yehuda just applied it to 2-3-stable. It's fixed in master as far as I know because rack's multipart params parsing is a lot smarter. A test case for it would be a really good idea though.

Actually the patch provided April 29th, 2009 @ 08:16 AM (https://rails.lighthouseapp.com/projects/8994/tickets/2576/a/116520...) contains a test. However the following diffs I have submitted is without this test.

The test probably needs to be updated to the latest 2-3-stable branch.

Why is this changed from 2.3.9 to 3.x?

Jarl

Jarl, it was applied on 2.3.9. We just need a test to be added to Rails 3.0 to ensure we won't have regressions.

@José: Thanks for the information, looking forward to 2.3.9. Still I hope you can use the test in the patch to make a test in 3.x, but it sounds like the architecture changes requires a new test to be written.

Why is state changed to stale? and what does it mean?

I'm marking it as stale as I didn't get a patch back. I will gladly reopen it if one is provided.

I have tested this in 2.3.9, and it seems fixed there.

I consider this bug as resolved, fix released.