www.libssh2.org | Daily snapshots | Mailing list archive | Docs | Examples | github

Archive Index This month's Index

Subject: Re: configure.ac crypto library checks [was: 1.8.0]

Re: configure.ac crypto library checks [was: 1.8.0]

From: Kamil Dudka <kdudka_at_redhat.com>
Date: Thu, 27 Oct 2016 13:40:18 +0200

On Thursday, October 27, 2016 12:44:47 Tor Arntsen wrote:
> On 27 October 2016 at 13:15, Peter Stuge <peter_at_stuge.se> wrote:
> > Kamil Dudka wrote:
> >> On Thursday, October 27, 2016 01:05:08 Peter Stuge wrote:
> >> > Also, I would like to argue that given --with-x but no x found should
> >> > be considered a fatal error by configure. Any objections to that?
> >>
> >> Sounds like a good idea. By fatal error, you mean to return non-zero
> >> exit
> >> status from configure script, or exit immediately and skip all the
> >> remaining checks?
> >>
> >> I would prefer the former.
> >
> > Right, certainly exit non-zero.
> >
> > I was thinking exiting then and there (with a useful error message)
> > when the requested x is not available.
> >
> > Why do you prefer running more checks when a failure can't be avoided
> > anyway?
>
> I would prefer the latter. It's messy to work with configure scripts
> which just continue after detecting something missing, with tons of
> more output, and then have to scroll back page after page to try to
> spot where it failed (or where it first failed). Better to exit
> directly there and then, the failed check right at the end, fix (e.g.
> install a missing library), run configure again, repeat as necessary.
> Much easier.

It sounds more like a reason to summarize all errors after the configure
script finishes, not a reason to skip the remaining configure checks. It
would make sense to skip them if the configure script ran overly long but
it is fortunately not the case with libssh2.

Kamil
_______________________________________________
libssh2-devel https://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel
Received on 2016-10-27

the libssh2 team