Re: rcs-5.8-1 on HP-UX 11.31

[Paul Ackersviller]

On Tue, Jul 10, 2012 at 12:09:31AM +0200, Thien-Thi Nguyen wrote:
> () Paul Ackersviller <address@hidden>
> () Mon, 9 Jul 2012 21:32:36 +0000
> 
>    I may have spoken a little bit too soon last time.  I'm getting
>    some strange behaviour that's hard to explain, but here's a try.
>    After an initial configure, the first make gets stuck several
>    time waiting for stdin from the terminal, i.e. on each of these
>    sed invocations:
> 
> Good, we're getting more specific (GNU vs POSIX sed).  However,
> not specific enough -- i don't see the invocations you mentioned.
> Did you forget to include an attachment?

D'oh!  Sorry, forgot to paste it in... here's my make output tail end:
        sed 's/SUB/ci/;s/VERSION/5.8.119-8540/'  > ci
        chmod +x ci
        sed 's/SUB/co/;s/VERSION/5.8.119-8540/'  > co
        chmod +x co
        sed 's/SUB/rcs/;s/VERSION/5.8.119-8540/'  > rcs
        chmod +x rcs
        sed 's/SUB/rcsclean/;s/VERSION/5.8.119-8540/'  > rcsclean
        chmod +x rcsclean
        sed 's/SUB/rcsdiff/;s/VERSION/5.8.119-8540/'  > rcsdiff
        chmod +x rcsdiff
        sed 's/SUB/rcsmerge/;s/VERSION/5.8.119-8540/'  > rcsmerge
        chmod +x rcsmerge
        sed 's/SUB/rlog/;s/VERSION/5.8.119-8540/'  > rlog
        chmod +x rlog
Making all in tests
No suffix list.
`all-am' is up to date.

>    Hitting EOF for each one gives me a bunch of empty files, of
>    course.  A make clean, followed by re-make clears this all up!
> 
> Hmm, sounds tricky to investigate.  Is this behavior reproducible?

Yes, and less mysterious than I first realized... I'm not even sure now
why I thought my first make after configuration was somehow different,
that's really not the case.

Comparing the gmake output shows that the sed's above are missing
their "sub.TEMPLATE" input file.  So I don't think sed itself is the
trouble, unless it's somehow involved with generating the Makefile's.

> I need to brush up on the differences between GNU and POSIX sed,
> and audit the codebase appropriately.  In theory, the specific
> failures mentioned above will be addressed by the audit (so it's
> not strictly necessary to dig them up), but i wouldn't mind seeing
> some concrete examples anyway.

Let me know if you need any more info.