On 12/07/26 7:50 pm, Andrew Lee wrote:
> On Sun, Jul 12, 2026 at 3:26 PM Nilesh Patra <[email protected]> wrote:
>>> "DefaultGODBUG" is found.
>> Got it. I've written a check and it emits the following for hugo:
>>
>> W: hugo: binary-contains-defaultgodebug-options 
>> asynctimerchan=1,containermaxprocs=0,cryptocustomrand=1 ... [/usr/bin/hugo]
>> be a bit better, please suggest if something else would sound better.
> 
> I would suggest to use `go-binary-contains-godebug-overrides` instead
> of `defaultgodebug-options` to avoid confusion.
> 
> The phrase "defaultgodebug" makes it sound like a standard go default,
> which is not what upstream intends.

I renamed it to "binary-contains-insecure-defaultgodebug-settings". Does that 
sound OK?

Override generally means maintainer overrides (than upstream or compiler gen), 
so I'd
rather not use that wording.

On 12/07/26 8:22 pm, Dr. Tobias Quathamer wrote:
> If the MR makes it into lintian now, I think that "info" or even
> "pedantic" might be a better choice, as long as Debian still uses
> GO111MODULE=off as default setting.

I'm planning to merge this, yes. I've changed the level to pedantic for now.

> When that has changed in dh-golang, the check you've written could be
> bumped in severity to a warning.
> 
> For the explanation, you might consider this text:
> 
> 
> The binary contains legacy or insecure DefaultGODEBUG variable settings.
> 
> Modern Go versions allow developers to force legacy (and potentially
> insecure) behaviors using //go:debug directives or GODEBUG settings.
> While these are sometimes necessary for temporary compatibility, they
> can introduce security vulnerabilities, disable critical fixes, or
> compromise build reproducibility.
Added. Could you please review the MR and comment on it, please?

> Am 12.07.26 um 15:26 schrieb Nilesh Patra:
>> Not sure how useful the individual options are, I wonder if it makes
>> sense to not emit them at all.
> Currently, the list is really long, yes. So it might make sense to just
> emit the tag without the specific DefaultGODEBUG settings.
> 
> However, when the go packages are build with GO111MODULE=on, that list
> should normally be empty. In that case, if there are only very few
> options listed, I think it would help the maintainer to see which
> options were used, in order to determine why those settings were used
> and if they are really needed.
> 
> My understanding is that the Debian archive would need binNMUs for all
> golang packages which produce an executable, to be rebuilt in a
> module-aware fashion. Afterwards, the list should be empty for almost
> all packages.

I just cut a release today, and lintian release happens roughly every month.
Given that some work on dh-golang has already happened, would 1 month be a fair
enough time for this?

Thanks,
Nilesh

Reply via email to