Re: [RFA] sed bug in Makefile.in

["Theodore A. Roth" <troth@openavr.org>]

On Fri, 11 Jul 2003, Daniel Jacobowitz wrote:

> On Fri, Jul 11, 2003 at 10:14:01AM -0700, Theodore A. Roth wrote:
> > Hi,
> >
> > I just tried to configure and build gdb with:
> >
> >   ../configure --prefix=$HOME/local/avr --program-suffix=-6.0 --target=avr --enable-gdb-build-warnings=,-Werror
> >   ../configure --prefix=$HOME/local/avr --program-suffix=-6.0 --program-prefix=avr- --target=avr --enable-gdb-build-warnings=,-Werror
> >
> > Both failed to install avr-gdb-6.0 which I expected.
> >
> > The failure (on 'make install') is this:
> >
> > make[1]: Leaving directory
> > `/home/roth/dev/tools/gdb-6.0-cvs/obj-avr/gdb'
> > transformed_name=`t='s,$,-6.0,; s,^,avr-,; '; \
> >                   echo gdb | sed -e $t` ; \
> >         if test "x$transformed_name" = x; then \
> >           transformed_name=gdb ; \
> >         else \
> >           true ; \
> >         fi ; \
> >         /bin/sh ../../gdb/../mkinstalldirs /home/roth/local/avr/bin ; \
> >         /usr/bin/install -c gdb \
> >                 /home/roth/local/avr/bin/$transformed_name ; \
> >         /bin/sh ../../gdb/../mkinstalldirs \
> >                 /home/roth/local/avr/man/man1 ; \
> >         /usr/bin/install -c -m 644 ../../gdb/gdb.1 \
> >                 /home/roth/local/avr/man/man1/$transformed_name.1
> > sed: can't read s,^,avr-,;: No such file or directory
> >
> >
> > The attached patch seems to fix this.
> >
> > Ok to commit? Should probably go on both head and branch.
>
> Yes, this is fine.

Thanks.

This only fixes the case when both --program-prefix and
--program-suffix are specified.

There still seems to be a problem with the case where --target and
--program-suffix are specified. The result with just --target=avr is
the installed program being named avr-gdb, but when using --target and
--program-suffix, the avr- prefix is lost. This seems to caused by
this code in configure:

   1148 test "$host_alias" != "$target_alias" &&
   1149   test "$program_prefix$program_suffix$program_transform_name" = \
   1150     NONENONEs,x,x, &&
   1151   program_prefix=${target_alias}-

So, it may not be fixable (or need to be fixed). I guess the behavior
just surprised me.

Ted Roth