Re: RFE: configure -> dependency list on exit.

[Hugh Sasse Staff Elec Eng]

On Fri, 4 Mar 2005, Stepan Kasal wrote:

> Hi,
>  thanks.  I'll make a few comments, but we should wait for Paul's

Well, you'd reduced it to something manageable, so thank you.

> opinion, it's more important than mine.
>
> On Fri, Mar 04, 2005 at 01:16:08PM +0000, Hugh Sasse Staff Elec Eng wrote:
> > +Display a message when @code{AC_OUTPUT} is called.
>
> The node describing AC_OUTPUT should be changed, too.  We should explain

I wondered about that, but wasn't sure if the repetition would be
helpful.

> that if there are messages pending from AC_MSG_END or AC_MSG_NEED,
> these messages are displayed, _insted of_ the normal procedure.

I'd have put _at the end of_ the normal precedure.  We may want to
know that some things weren't available and things would work better
if we had them, but one can still get useful functionality
nonetheless, so building is OK.  Can't think of any (good) examples,
except GNU make, GNU diff do more than many native versions.
Maybe AC_MSG_NEED([[Gnu Diff]],
[[To get unified context diffs you can get
ftp://ftp.gnu.org/gnu/diff]])


> > address@hidden AC_MSG_NEED (@var{dependency-description}, @ovar{priority})
>
> My idea was to have AC_MSG_NEED with args
>   (@var{dependency-name}, @var{dependency-text}, @ovar{priority})
> used like
>
> AC_MSG_NEED([flex],
>  [Get the latest version of lex from ftp://ftp.gnu.org/gnu/flex/.])

Yes, that's better. Sorry, I'll fix that.  I wrote it from memory
rather than going back and picking all the points out.

Should dependency-name be a package or a desired feature, which the
suggested package(s) support?  There are more kinds of awk than I
care to think about, for example.  Maybe any one would do, so
insisting on GNU awk may be too pedantic when its probably on an
options CD for some OS.  [No, I can't think of an example OS for
this awk case, but...]
        [...]
> I thought that having two parameters, NAME and TEXT, could help some
> tools which would trace all AC_MSG_NEED calls, in order to get the list
> of all requiremets for the package.

And probably worth mentioning that in the text.
> But I think Paul didn't like it.

I tried to base AC_MSG_NEED on his more general idea, but I think
the readability is good (as you saw from the text).
> > +AC_MSG_NEED([A lexical analyser is needed by Bison, such as Flex or
> > +lex.])
> > address@hidden Not sure if there are sufficient []'s there....
>
> Yes, this is correct.
+AC_MSG_NEED([[A lexical analyser is needed by Bison, such as Flex or
+lex.]])
???
> > address@hidden AC_MSG_END (@var{message}, @ovar{priority})
>
> I think Paul's idea was to haves
>          AC_MSG_NOTICE (@var{message}, @ovar{priority})
> if the second argument is omited, the message is displayed immediately.
> Otherwise, the message is saved for the end.

I think that is harder to understand, and this would be clearer when
reading.  For example, when hacking the texi it was pretty clear
what the @code{}, @file{} etc things did.
> Thank you for the proposal.  Let's see the reactions.
>
> Stepan
        Thank you,
        Hugh