Re: [PATCH] Replacement for the sigs_to_ignore hack in timeout.c

[Jim Meyering]

Giuseppe Scrivano <address@hidden> wrote:
> do you have suggestions on this patch?  It replaces any `signal' with
> `sigaction'.
>
> Regards,
> Giuseppe Scrivano
>
> Jim Meyering <address@hidden> writes:
>
>> Good idea.
>> I've wanted to get rid of "signal" uses for ages.
>> Are you interested in doing it?

Giuseppe,

Perhaps you misunderstood?
Eric proposed that idea and I addressed the reply above to him.
He then replied that he would indeed like to implement it.

Since you have spent time working on it, in spite of that,
we are now in a bind.

  - you don't have a copyright assignment on file for coreutils,
    and those take at least 2-3 weeks to process.  So even if we
    were to use your work, we'd have to wait.

  - it is considered poor form to jump in and write/submit a patch
    like that, when someone else is obviously planning to do the
    same thing.

I suggest that Eric go ahead and do what he offered to,
and post that.  Then, perhaps it'd be interesting to compare the
two patches, but I expect to review and use his.

If you'd like to contribute, please coordinate, especially if
it's a task that's being actively discussed.

If you'd like another idea, here's a task I'd like to see completed,
but for which no one has volunteered:

[excerpt from
http://thread.gmane.org/gmane.comp.gnu.coreutils.bugs/14417/focus=14546 ]

    [dd already has your "b =512, ..." notation, but 8 other programs
     with the same strings do not]

    I've applied this, but omitted the above part, because it adds "="
    marks to just one of the several files that would need it in order to
    keep those duplicated messages in sync.  However, I would welcome a patch
    that unifies and (if possible) then factors out some of the duplication
    displayed by this command:

        grep -E 'MB? =?10..\*10' src/*.c

    Ideally, each message string would appear in just one place,
    and all clients would use some symbol from e.g., system.h.