Add linter warnings for missing ownership annotations

[Ayushd785]

description

adds a linter warning for functions that return ownership-sensitive pointer types (RzList, RzPVector, RzGraph) without an explicit ownership annotation.

When a function returns one of these types as a pointer and is missing RZ_OWN or RZ_BORROW, the linter now emits a warning indicating the function and location.

ScreenShot

Fixes
#34

[Rot127]

Like the change. But why only these types? Any pointer which is not const needs ownership info.

[Ayushd785]

Yaaa makes sense, this should apply to all non-const pointers. I'll update the logic to check if the return type is a non-const pointer instead of checking against specific type names. Will it be a good approach then?

[kazarmy]

const merely means that the data structure and/or pointer is read-only ... the caller might still need to free the returned const pointer depending on whether the annotation is RZ_OWN or RZ_BORROW

[Ayushd785]

Hmm, so you're saying even const pointers need ownership annotations since const only affects mutability, not ownership?
But then, should we check ALL pointer returns, or still focus on specific types like container types (RzList, RzVector, etc.) that are heap-allocated?

[wargio]

const merely means that the data structure and/or pointer is read-only

i disagree. i expect any const pointer (returned) to be always borrowed

[kazarmy]

i expect any const pointer (returned) to be always borrowed

That is a reasonable expectation and should be coded in the linter logic, meaning that there should be a warning when RZ_OWN is used with a returned const pointer.

Hmm, still need to check all pointer returns.

Anyway, I'm ok with checking all non-const pointer returns as a start. Hopefully there won't be a ton of warnings.

[notxvilka]

Yes, lets do all non-const pointers. I assume the amount of changes would be significant. We can run linter manually first before merging, fix all occurences, and then merge the PR

[Rot127]

meaning that there should be a warning when RZ_OWN is used with a returned const pointer.

Good idea! Second this.

[Ayushd785]

Got it! So i have to check ALL non-const pointer returns for missing ownership annotations.

[Ayushd785]

I am sorry, everyone, for being inactive here. I am on a very tight schedule for the next 3-4 days. I will implement the suggested changes as soon as i get some time.

[notxvilka]

It works! up for any feedback Thank you!

You will need excluding the subprojects/ that are external (like pcre2 on your screen) and not provided by Rizin itself (like rzheap).

[Ayushd785]

It works! up for any feedback Thank you!

You will need excluding the subprojects/ that are external (like pcre2 on your screen) and not provided by Rizin itself (like rzheap).

Okay i'll do it

[Ayushd785]

It works! up for any feedback Thank you!

You will need excluding the subprojects/ that are external (like pcre2 on your screen) and not provided by Rizin itself (like rzheap).

Ahhh! The bug was here

for command in commands:
            abspath = os.path.abspath(
                os.path.join(command["directory"], command["file"])
            )
            relpath = os.path.relpath(abspath, rizin_path)

            if relpath.startswith("subproject") or relpath.startswith("test"):
                continue

There is a bug it should be subprojects not subproject. That's why the check is ignored. I'll fix that

[Ayushd785]

hi @notxvilka, I've added an extra check in check_translation_unit() to
filter out external subproject headers

if "/subprojects/" in abspath:
            subproject_name = abspath.split("/subprojects/")[1].split("/")[0]
            if not subproject_name.startswith("rz"):
                continue

This assumes anything in subprojects/ that doesn't start with "rz" is an
external library and skips the linter check on it.

Also fixed that minor typo.
I am up for any feedback. Thank you!

[Ayushd785]

wait I have missed something. I need a logic which includes everything which has initials as 'rz' but does not have an underScore with it (eg. rz_util) i thought it was just rzheap

edit - fixed

[Ayushd785]

Screencast.From.2026-02-01.00-55-12.mp4

[notxvilka]

Ok, this is ready now. But before merging this, we need to fix those warnings in Rizin first, so that CI will not become instantly red.

[Ayushd785]

Ok, this is ready now. But before merging this, we need to fix those warnings in Rizin first, so that CI will not become instantly red.

Okay Sure.

[Ayushd785]

Ok, this is ready now. But before merging this, we need to fix those warnings in Rizin first, so that CI will not become instantly red.

Well, I was just wondering. what approach are you planning to use to fix
those thousands of linter warnings? I am just curious to know how you gonna automate it.?

[Rot127]

lgtm. But yeah, fixing all the warnings is needs some smart idea.

[notxvilka]

Created the issue to track this: rizinorg/rizin#5870

[Rot127]

What about we add a file in our repo root which contains the modules to check?

In the beginning it contains only modules which don't throw any warnings.
Then, we can fix up each module, and add its name in the file so rz-bindgen checks it?

Otherwise I don't see us making any progress?

[Ayushd785]

What about we add a file in our repo root which contains the modules to check?

In the beginning it contains only modules which don't throw any warnings. Then, we can fix up each module, and add its name in the file so rz-bindgen checks it?

Otherwise I don't see us making any progress?

This would be a great approach! CI stays green while we make progress. In addition to this We could also add a script that auto-detects which modules are already clean (0 warnings) and suggests adding them to the whitelist. I'd be happy to implement this if it's useful!