Re: YYSTYPE_IS_TRIVIAL

[Akim Demaille]

| > From: Akim Demaille <address@hidden>
| > Date: 24 Apr 2002 11:43:11 +0200
| > 
| > | I currently get the following in my attempt to bootstrap:
| > | 
| > | 433-sic $ ./bootstrap
| > | Bootstrapping CVS Bison...
| > | autoreconf: working in `.'
| > | autoreconf: running: aclocal -I m4 --output=aclocal.m4t
| > | autoreconf: `aclocal.m4' is created
| > | autoreconf: running: gettextize --copy --force
| > | Wiping out intl/ subdirectory
| > 
| > I really have absolutely no idea of what may cause this.  I cannot
| > reproduce it.
| 
| So, what you're saying is that "gettextize --copy --force" should not
| say "Wiping out intl/ subdirectory"?  But that is the normal thing for
| gettextize 0.11.1 to do if there is an intl subdirectory.  It wipes
| out everything but RCS or CVS subdirectories.

Actually, it looks like I forgot a part of my sentence: how come
--intl was not passed.  Is it possible that your path didn't have
Gettext 0.11 at that time?


| > I have changed the logic of bootstrap (which is necessarily a bit
| > weird, since we do expect a failure of autoreconf due to gettextize).
| > I think it is more robust now:
| 
| OK, here's the output of "./bootstrap; ./configure; make check" on my
| platform, using the latest CVS.  This uses Solaris 8, Autoconf 1.53a,
| Automake 1.6.1, and gettext 0.11.1.  Is this the behavior that you
| expect?  If not, perhaps you could send an email containing what
| "./bootstrap; ./configure; make" outputs on your host?

What you show is what I expect.

Except (hm, nice word after the previous one :) the following, of course:


| running: /opt/sfw/bin/gm4 -I /net/knick/home/eggert/src/gnu/bison.tmp/data 
m4sugar/m4sugar.m4 /tmp/bsnWca4qC bison.simple
| sets.at:82: sed -f extract.sed stderr
| --- /dev/null 2002-04-24 10:43:19.000000000 -0700
| +++ /net/knick/home/eggert/src/gnu/bison.tmp/tests/testsuite.dir/at-stderr    
2002-04-24 10:43:20.852402000 -0700
| @@ -0,0 +1 @@
| +Unrecognized command:    /^[         ]*$/! b null
| stdout:

This is sed talking.  The program is in sets.at:

# AT_EXTRACT_SETS(INPUT, OUTPUT)
# ------------------------------
# Extract the information about the grammar sets from a bison
# trace output (INPUT), and save it in OUTPUT.
# And remember, there is no alternation in portable sed.
m4_define([AT_EXTRACT_SETS],
[AT_DATA([extract.sed],
[[#n
/^NULLABLE$/ {
   :null
   p
   n
   /^[  ]*$/! b null
}
/^FIRSTS$/ {
   :firsts
   p
   n
   /^[  ]*$/! b firsts
}
/^FDERIVES$/ {
   :fderiv
   p
   n
   /^[  ]*$/! b fderiv
}
/^DERIVES$/ {
   :deriv
   p
   n
   /^[  ]*$/! b deriv
}
]])
AT_CHECK([sed -f extract.sed $1], 0, [stdout])
AT_CHECK([mv stdout $2])
])


Apparently, it contains nonportable sed, but I don't know where :(

-- 
Ashamed.