Backports of foreign patches?

Backports of foreign patches?

[Harald Anlauf]

Dear all,

just a hypothetical question: assume that some patch for a regression was
applied only to e.g. master some time ago but might be worth a backport,
the regression was indicated in the PR, and the patch author does not answer
any related queries in the PR?

How could/should/would one proceed?

Would it be acceptable to backport the patch after a sufficient waiting time
(at least e.g. one or two months)?  With a notification to the ML?  Or does
it require an additional OK from a reviewer?

Or would one pass an invisible red line?

Thanks,
Harald

Re: Backports of foreign patches?

[Thomas Koenig]

Hi Harald,

What I'm writing below is my personal opinion, if anybody feels
differently, please chime in.

 > just a hypothetical question: assume that some patch for a regression was
 > applied only to e.g. master some time ago but might be worth a backport,
 > the regression was indicated in the PR, and the patch author does not 
answer
 > any related queries in the PR?
 >
 > How could/should/would one proceed?
 >
 > Would it be acceptable to backport the patch after a sufficient 
waiting time
 > (at least e.g. one or two months)?

I would say yes.  Using "git gcc-backport" automatically credits
the original author, so that's OK.  Regarding the time frame, it
probably makes sense to backport before the next release of that
particular branch.

That is, if it does apply cleanly.  If there is some work needed
that exceeds being cosmetic, then this would in effect be a new
patch, which would require review.

 > With a notification to the ML?

Not needed if it applies cleanly.

 > Or does
 > it require an additional OK from a reviewer?

Only if it is a substantial rewrite (see above).

 > Or would one pass an invisible red line?

Certainly not for me - I would feel bad about not reacting
in the first place, but not about somebody backporting
one of my patches.

Best regards

     Thomas

Re: Backports of foreign patches?

[Paul Richard Thomas]

Hi Harald,

I agree completely with Thomas on this. That said, don't be too diligent
about it. The gcc version with Fedora 33 is 10.2.1 and I presume that other
distros are keeping similarly up to date.

Regards

Paul


On Wed, 30 Dec 2020 at 09:53, Thomas Koenig via Fortran <fortran@gcc.gnu.org>
wrote:

> Hi Harald,
>
> What I'm writing below is my personal opinion, if anybody feels
> differently, please chime in.
>
>  > just a hypothetical question: assume that some patch for a regression
> was
>  > applied only to e.g. master some time ago but might be worth a backport,
>  > the regression was indicated in the PR, and the patch author does not
> answer
>  > any related queries in the PR?
>  >
>  > How could/should/would one proceed?
>  >
>  > Would it be acceptable to backport the patch after a sufficient
> waiting time
>  > (at least e.g. one or two months)?
>
> I would say yes.  Using "git gcc-backport" automatically credits
> the original author, so that's OK.  Regarding the time frame, it
> probably makes sense to backport before the next release of that
> particular branch.
>
> That is, if it does apply cleanly.  If there is some work needed
> that exceeds being cosmetic, then this would in effect be a new
> patch, which would require review.
>
>  > With a notification to the ML?
>
> Not needed if it applies cleanly.
>
>  > Or does
>  > it require an additional OK from a reviewer?
>
> Only if it is a substantial rewrite (see above).
>
>  > Or would one pass an invisible red line?
>
> Certainly not for me - I would feel bad about not reacting
> in the first place, but not about somebody backporting
> one of my patches.
>
> Best regards
>
>      Thomas
>


-- 
"If you can't explain it simply, you don't understand it well enough" -
Albert Einstein

Fw: Re: Backports of foreign patches?

[Harald Anlauf]

I'm forwarding this to the ML, as I did not pay sufficient attention
to the distribution list...

> Gesendet: Mittwoch, 30. Dezember 2020 um 12:11 Uhr
> Von: "Harald Anlauf"
> An: "Thomas Koenig"
> Betreff: Re: Backports of foreign patches?
>
> Hi Thomas, Paul,
>
> thanks for the clear statements.
>
> I am less concerned about what I have at home than what
> I might have supported at work say two or three years from now.
>
> We now have ("just got")
>
> % cat /etc/redhat-release
> Red Hat Enterprise Linux ComputeNode release 7.7 (Maipo)
>
> which has a baseline of
>
> gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-39)
>
> but for real work also have an installed module with 9.1.0.
> Expectations might be more realistic for a higher minor version
> than for a higher major version...
>
> Best regards,
> Harald
>