Update of sr #111044 (group autoconf):
Status: Need Info => Confirmed
_______________________________________________________
Follow-up Comment #4:
Copying conversation from bug-autoconf:
Frank Eigler:
> On the other hand, we've had very real reports that e.g. diff was
> missing on some real platforms, which led autoconf astray and ended up
> in misconfigured build trees. Perhaps autoconf can afford to do some
> sanity checking on these reasonable-sounding but not-universal
> expectations.
Zack Weinberg:
> Frank, do you have details of those real platforms with missing diff?
> As I wrote in <https://savannah.gnu.org/support/index.php?111044>, it's
> difficult to implement this kind of sanity check without access to an
> environment that *fails* this kind of sanity check. I could set one up
> myself but I wouldn't know if I had picked the right things to leave out.
Frank Eigler:
> Not a live system I use day-to-day. But I mean I can remove
> "diffutils" from my Fedora box, and see it break.
>
> I think the point was not so much that normal & typical developer
> systems may fail. It's that unforseen restricted machines (maybe
> running in minimal fashionably-"secure" buildroots) could encounter
> these problems, and then QUIETLY FAIL by misconfiguring tools.
>
> Thus the idea was just to do lightweight & obvious sanity checking on
> the environment at configure startup. e.g.:
>
> diff /dev/null /dev/null || exit 1
I think this gives me enough information to proceed.
_______________________________________________________
Reply to this item at:
<https://savannah.gnu.org/support/?111044>
_______________________________________________
Message sent via Savannah
https://savannah.gnu.org/
