Bug 894684 – dev-libs/cyrus-sasl: minor compilation issue of srp plugin (plugins/srp_init.c:41:1: warning: type of 'srp_client_plug_init' does not match original declaration [-Wlto-type-mismatch])

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description Jocelyn Mayer 2023-02-16 06:49:19 UTC]

Created attachment 851510 [details, diff]
cyrus-sasl srp plugin patch to avoid the described warnings

The src_client_plug_init / src_server_plug_init functions implementations in plugins/srp.c have a last "plugname" unused argument which is not in the prototype generated by the SASL_CLIENT_PLUG_INIT / SASL_SERVER_PLUG_INIT macros from common/plugin_common.h
This leads to 2 compilations warnings:
cyrus-sasl-2.1.28/plugins/srp_init.c:41:1:
warning: type of 'srp_client_plug_init' does not match original declaration [-Wlto-type-mismatch]
cyrus-sasl-2.1.28/plugins/srp_init.c:42:1:
warning: type of 'srp_server_plug_init' does not match original declaration [-Wlto-type-mismatch]

This can easily be avoided by removing the plugname argument (patch following).
As this minor issue seems to still be present in trunk code, a ticket has been opened at cyrus-sasl repository:
https://github.com/cyrusimap/cyrus-sasl/issues/757

[Comment 1 Sam James 2023-02-16 07:05:53 UTC]

Good spot. Thanks for filing these, code quality is important.

FWIW: for the upstream reports, most projects will prefer a PR rather than an inline patch in a bug, but if you're giving a patch in a bug upstream (no need if attaching), make sure to use Markdown so it renders ok, like this:
```
--- a/foo.c
+++ b/foo.c
...
```

(it's just 3 backticks above/below the 'code' bit)

[Comment 2 Sam James 2023-02-16 07:07:34 UTC]

It's possible that the unused int is there for ABI reasons (so they can add something later without breaking ABI), but it's a bit weird. So they might prefer to do something else to fix the warning.

[Comment 3 Jocelyn Mayer 2023-02-16 19:18:21 UTC]

(In reply to Sam James from comment #1)
> Good spot. Thanks for filing these, code quality is important.
> 
> FWIW: for the upstream reports, most projects will prefer a PR rather than
> an inline patch in a bug, but if you're giving a patch in a bug upstream (no
> need if attaching), make sure to use Markdown so it renders ok, like this:
> ```
> --- a/foo.c
> +++ b/foo.c
> ...
> ```
> 
> (it's just 3 backticks above/below the 'code' bit)

oh, OK, I did not check the view in the upstream ticket...
Just updated the inline patch upstream...

[Comment 4 Jocelyn Mayer 2023-02-16 19:22:46 UTC]

(In reply to Sam James from comment #2)
> It's possible that the unused int is there for ABI reasons (so they can add
> something later without breaking ABI), but it's a bit weird. So they might
> prefer to do something else to fix the warning.

it does not seem to me that this is the case as the macro used to define the headers declaration for plugins init routine does not declare this argument.
Then, using it from a caller would lead to some kind of issues, I guess.
And, as it seems, other plugins using this macro do not implement this extra argument...
Upstream will decide the best way to fix it...

[Comment 5 Larry the Git Cow 2024-04-26 12:49:59 UTC]

The bug has been closed via the following commit(s):

https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=9f10d53dc71950701eb206c84c62019ac0fe44ca

commit 9f10d53dc71950701eb206c84c62019ac0fe44ca
Author:     Eli Schwartz <[e-mail]>
AuthorDate: 2024-04-26 05:18:38 +0000
Commit:     Sam James <[e-mail]>
CommitDate: 2024-04-26 12:48:54 +0000

    dev-libs/cyrus-sasl: mark as LTO-unsafe
    
    Fixed upstream in git master.
    
    Closes: https://bugs.gentoo.org/894684
    Signed-off-by: Eli Schwartz <[e-mail]>
    Signed-off-by: Sam James <[e-mail]>

 dev-libs/cyrus-sasl/cyrus-sasl-2.1.28-r4.ebuild | 7 +++++++
 1 file changed, 7 insertions(+)