bug#17246: grep-2.19 planning

[Paolo Bonzini]

Il 12/04/2014 19:35, Paul Eggert ha scritto:
> Paolo Bonzini wrote:
> > It is also a regression that error messages will show "grep: error xyz"
> > rather than "egrep: error xyz".
>
> That's a feature, not a bug :-).  It reminds people to use 'grep' rather
> than an obsolete extension.

Ok, I can buy this. :)

> > Please provide a testcase for your bug, I'm not sure whether I
> > understand it.
>
> Ouch, in looking back at it I now see that *I* misunderstood it.  You're
> correct: that "if" statement avoided the problem that I saw.  The point
> remains that the previous code was Too Complicated, though.
>
> Perhaps in penance I should write the little C wrappers I was talking
> about.

C wrappers are better than shell scripts in terms of 
portability---remember that Windows does have decent shells such as 
PowerShell, just not POSIX ones.  However, I'm not sure that the 
complexity of C programs will be smaller than the complexity of grep 
<=2.18's machinery.

One remaining problem with C wrappers is that they make the grep 
installation non-relocatable and "non-separable".  These are two 
features that I'm afraid people are probably relying on for Windows targets.

There is certainly room for deprecating and/or removing part of the 
"matcher" machinery.  Why can't we take the safest path and revert for 
2.19, and afterwards work on cleanups?

Paolo