git magic or usage wisdom.

git magic or usage wisdom.

[jim.cromie]

so I have this patchset (sent to lkml recently ),
it adds a new struct:
struct _ddebug_info

dyndbg: create and use struct _ddebug_info

    this new struct gathers the linker provided vectors/sections:

         descs - the vector of descriptors in __dyndbg section.
         num_descs - length of the data/section



If I had an endless supply of free do-overs,
Id change the name to something a little less user-facing,
since its meant to ref the whole dyndbg-config-state

is there some git command magic to work this ?

in a pre-git world, I might try
perl -pi -e 's/_ddebug_info/_ddebug_stateinfo/g' 00*.patch

that "might" work, but could also create a myriad of conflicts
when the patchset is 'git am' d
and it might apply clean but still break on compile ?

anyone care to opine on the probability of success ?

If I try this, I'll report back.
ISTM more likely than my doing it manually.

_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

Re: git magic or usage wisdom.

[Ismael Luceno]

On 04/Aug/2022 13:24, jim.cromie@gmail.com wrote:
> so I have this patchset (sent to lkml recently ),
> it adds a new struct:
> struct _ddebug_info
<...> 
> is there some git command magic to work this ?
> 
> in a pre-git world, I might try
> perl -pi -e 's/_ddebug_info/_ddebug_stateinfo/g' 00*.patch
> 
> that "might" work, but could also create a myriad of conflicts
> when the patchset is 'git am' d
> and it might apply clean but still break on compile ?
> 
> anyone care to opine on the probability of success ?
> 
> If I try this, I'll report back.
> ISTM more likely than my doing it manually.

The approach is fine :).

Whatever you do, you can't escape testing that the patches apply and
compile...

What you may simplify is the editing by chaining the commands to
replay changes into a new branch:

	git format-patch -k --stdout base..old-branch |
	  perl -pe '...' | git am -3 -k

Further automation seems counterproductive.

As for the editing, I would use the following perl code:

	if (/^\+/) { s/\b_ddebug_info\b/_ddebug_stateinfo/g }

That way you limit edits only to new content, and the struct name is
properly bounded by the regex.

I guess that's about as robust as you can make it without going out of
your way.

_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

Re: git magic or usage wisdom.

[jim.cromie]

On Thu, Aug 4, 2022 at 1:58 PM Ismael Luceno <ismael@iodev.co.uk> wrote:
>
> On 04/Aug/2022 13:24, jim.cromie@gmail.com wrote:
> > so I have this patchset (sent to lkml recently ),
> > it adds a new struct:
> > struct _ddebug_info
> <...>
> > is there some git command magic to work this ?
> >
> > in a pre-git world, I might try
> > perl -pi -e 's/_ddebug_info/_ddebug_stateinfo/g' 00*.patch
> >
> > that "might" work, but could also create a myriad of conflicts
> > when the patchset is 'git am' d
> > and it might apply clean but still break on compile ?
> >
> > anyone care to opine on the probability of success ?
> >
> > If I try this, I'll report back.
> > ISTM more likely than my doing it manually.
>
> The approach is fine :).
>

thanks for the confirm, and I like the ^+ filter, and the \b's
it limits the modifications tightly.

> Whatever you do, you can't escape testing that the patches apply and
> compile...
>

indeed. its embarrassing when crap escapes the lab.

> What you may simplify is the editing by chaining the commands to
> replay changes into a new branch:
>
>         git format-patch -k --stdout base..old-branch |
>           perl -pe '...' | git am -3 -k
>
> Further automation seems counterproductive.
>

that pipeline is helpful, I will play with it.


> As for the editing, I would use the following perl code:
>
>         if (/^\+/) { s/\b_ddebug_info\b/_ddebug_stateinfo/g }
>
> That way you limit edits only to new content, and the struct name is
> properly bounded by the regex.
>
> I guess that's about as robust as you can make it without going out of
> your way.

_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies