asynchttpserver form data/body broken with #13147

[zedeus]

#13147 claims it does not break anything (and states it's an option), but it broke Jester's form data handling. Since there is a transition from body to bodyStream, this might have been intended.

import jester
routes:
  post "/":
    resp $request.params

nim c -d:useStdLib test.nim
curl -d 'q=foo' http://localhost:5000

With devel: {:}
Before the change: {"q": "foo"}

The Jester code responsible for getting the form data is not easily modifiable to support this change as it's non-async and relies on the old body field which is now useless.

[ghost]

But #13146 wasn't merged (it's closed)?
Maybe you mean https://github.com/nim-lang/Nim/pull/13339/files#diff-3133183ae7c444b928e3e590f4b1d217R222 ?

[zedeus]

Sorry, relied on Github's inline preview menu and didn't notice the duplicate.

[mrhdias]

#13147 claims it does not break anything (and states it's an option), but it broke Jester's form data handling. Since there is a transition from body to bodyStream, this might have been intended.

import jester
routes:
  post "/":
    resp $request.params

nim c -d:useStdLib test.nim
curl -d 'q=foo' http://localhost:5000

With devel: {:}
Before the change: {"q": "foo"}

The Jester code responsible for getting the form data is not easily modifiable to support this change as it's non-async and relies on the old body field which is now useless.

Sorry, but, my first proposal didn't break anything, you can see the code here: . 4fdbf8f

But Dominik Picheta asked me if I could implement it with Future Streams. Yes, with Future Streams it break the things. But, now you can upload large files without exhaust all resources ;-)

[andreaferretti]

Rosencrantz was also broken recently, I am going to check if this is the reason

[mrhdias]

#13147 claims it does not break anything (and states it's an option), but it broke Jester's form data handling. Since there is a transition from body to bodyStream, this might have been intended.

import jester
routes:
  post "/":
    resp $request.params

nim c -d:useStdLib test.nim
curl -d 'q=foo' http://localhost:5000

With devel: {:}
Before the change: {"q": "foo"}

The Jester code responsible for getting the form data is not easily modifiable to support this change as it's non-async and relies on the old body field which is now useless.

[andreaferretti]

I confirm that #13147 also broke Rosencrantz. It passes all tests on bfe96e0, but crashes starting from 955465e, which is the merge commit.

I will have a look and see if I can fix them, but wasn't there a promise not break things randomly after 1.0?

[Araq]

I will have a look and see if I can fix them, but wasn't there a promise not break things randomly after 1.0?

Actually the promise is two-fold:

1. Your code will continue to work with nim --useVersion:1.0.
2. Your code has a >95% chance of continuing to work without the --useVersion emulation switch.

And yeah, it means we need to revert the PR that introduced the problem.

[andreaferretti]

Your code will continue to work with nim --useVersion:1.0.

Cool! I had missed that, and in fact Rosencrantz still works with that flag :-)

[mrhdias]

I will have a look and see if I can fix them, but wasn't there a promise not break things randomly after 1.0?

Actually the promise is two-fold:

1. Your code **will** continue to work with `nim --useVersion:1.0`.

2. Your code has a >95% chance of continuing to work without the --useVersion emulation switch.

And yeah, it means we need to revert the PR that introduced the problem.

Please, check the fix for this issue: #13394

[ringabout]

#13147 was reverted by ec8a17c