Unknown type cowboy_router:route_match

[paulo-ferraz-oliveira]

Hi.

I'm running cowboy_swagger 1.2.3 with cowboy 1.1.2.

Static (dialyzer) analysis shows that cowboy_swagger trails.erl uses type cowboy_router:route_match(), which is not exported from cowboy 1.1.2's cowboy_router.

I don't know what cowboy version cowboy_swagger assumes for -spec().ing purposes, but I supposed (probably wrongly) that 2.x from cowboy_swagger would use 2.x from cowboy, for which the previous would be incompatible.

Would you be open to a support patch for 1.2.3 that fixes that issue?

Thanks.

[elbrujohalcon]

We sure am. Send that PR our way, @paulo-ferraz-oliveira !! Thank you in advance :)

[paulo-ferraz-oliveira]

I'd require a support branch (in the sense of git-flow, something like support/1.x) from c1258a1, though. Is that possible?

I'll do static analysis on 2.x to see if it suffers from a similar issue.

[elbrujohalcon]

Let me add you as a collaborator to this repo…

[elbrujohalcon]

Done! Go ahead and create branches! :)

[paulo-ferraz-oliveira]

👍 thanks.

I'm passing xref and dialyzer on both cowboy_trails and cowboy_swagger for OTP 19.3, 20.3 and 21.2. I (locally) changed rebar.config a bit (in both projects) to have my usual xref and dialyzer options. Will post results here, soon.

[paulo-ferraz-oliveira]

cowboy_trails

• with rebar3 xref, shows trails_handler: 1 unresolved call (I've never seen this before!),
• on OTP 21.2, has multiple compile-time warnings, related to cowboy and ranch,
• does not show other xref warnings,
• does not show other dialyzer warnings,
• with the remote rebar.config there's no new information

[elbrujohalcon]

unresolved call means "there is a dynamic call in there". Likely this one, which is fine since it is a behavior after all.

[paulo-ferraz-oliveira]

unresolved call means "there is a dynamic call in there". Likely this one, which is fine since it is a behavior after all.

But I've had dynamic calls in the past (for behaviours and not only) and I've never seen this warning.

[elbrujohalcon]

No idea, but that's also the only line of actual code in that module anyway. 🤷‍♂️

[paulo-ferraz-oliveira]

cowboy_swagger

• has a deprecation warning, related to cowlib,
• on OTP 21.2, has multiple compile-time warnings, related to cowboy and ranch,
• does not show other xref warnings,
• does not show other dialyzer warnings,
• with the remote rebar.config we get a bunch of unused exports (from xref) but that's fine!

[paulo-ferraz-oliveira]

combo cowboy master / cowboy_swagger master

• shows no warnings for xref and dialyzer, with OTP 19.3, 20.3 or 20.2 [I'm using 19+ because cowboy 2.x dropped support for 18.x], which means that once I migrate I won't have an issue anymore :D...

... but until I do, I'd like to PR trails (sorry, I though the issue was in cowboy_swagger, since I don't use trails directly).

[elbrujohalcon]

I'll add you as a collaborator there, as well.

[elbrujohalcon]

Done! :)

[paulo-ferraz-oliveira]

I can't seem to be able (403) to push to origin/upstream.

[paulo-ferraz-oliveira]

This is what I'm proposing for support/0.x (for trails): https://github.com/paulo-ferraz-oliveira/cowboy-trails/tree/support/0.x

[paulo-ferraz-oliveira]

No idea, but that's also the only line of actual code in that module anyway. 🤷‍♂️

It's due to xref_warnings. I'd never used it before.

[elbrujohalcon]

Cool. You should accept your invite at https://github.com/inaka/cowboy-trails/invitations

[elbrujohalcon]

This changes look good to me, BTW.

[paulo-ferraz-oliveira]

Here's what I'll propose for support/1.x: https://github.com/paulo-ferraz-oliveira/cowboy_swagger/tree/support/1.x

I need a minor rebar.lock update as soon as trails is hex-published, though.

[elbrujohalcon]

…and now it's published :)

[paulo-ferraz-oliveira]

@elbrujohalcon: do you think we can/should close this thread?

[elbrujohalcon]

Closing it!