Better error reporting about relm HOT 13 CLOSED

To have that right now, I could implement something that will run when an environment variable is set. It will do the following:

• Try to compile the source code.
• If there's an error within a #[widget] attribute:
  • Generate the source code annotated by this attribute and compile it.
  • Match the errors of the generated code with the original one.
  • Show these errors instead of the one produced by the compiler.

That will be slow, but still faster that doing that by hand.

from relm.

This is really needed, especially for noobs like me in Rust and you get this kind of error:

error: borrowed value does not live long enough
  --> src\widgets\daywidget.rs:34:9
   |
34 | #[widget]
   | --------^
   | |       |
   | |       temporary value dropped here while still borrowed
   | temporary value created here
   |
   = note: values in a scope are dropped in the opposite order they are created

I think I'll give up on relm for the moment :(

from relm.

I think that I'll soon be able to fix this issue (at least on nightly) since syn was fixed.
However, I think I need this pull request to be merged, but hopefully it will happen soon.

Until then, you can enable the log when you use relm by doing:

export RUST_LOG=warn

and you'll see the generated code.
Then, you can try the generated code to have a better error but it is annoying to do that.

from relm.

Oh cool, I'm gonna give this a shot to see if I can understand which value causes this error :)

Thank you. And to be fair, I know the whole Rust sphere is very young and still craving its way to provide all the tools needed to make a language successful :) So kudo to you for relm, it makes gtk way less ugly to use in Rust!

from relm.

Oh, I've just noticed that this logging feature is only in the feature/futures-glib branch (which contains many breaking changes).
Do you want that I add it in the master branch?

from relm.

If the PR needed to makes thing right by managing the current ticket is merged soon, don't bother, I can wait :)

from relm.

Looks like the pull-request was merged. Is it something we need to update on our end, or does the library need to be updated?

from relm.

@superlou: Unfortunately, this PR was merged with a hack, so it does not work right now.
Also, that will need to update the code to switch to proc-macro2.

from relm.

@zmitchell I talked to you about the idea of using (sy)nom, but I'm not sure we can use it for the whole thing.
I think we will only be able to parse the content of the view! macro with it.

from relm.

I agree, I think it's the only part that it needs to be used for. I don't think I implied otherwise, did I?

from relm.

@zmitchell No, but I did :) .

from relm.

@zmitchell: I started to fix this and have some basic examples (like buttons-attribute) working.
You can see my work in branch fix/better-error (see this giant commit).
Hopefully, I'll finish the work in 2-3 weeks.

from relm.

Fixed by PR #90.

from relm.