Re: [PATCH] Fix problem with generation of man/poke.1

[Jose E. Marchesi]

Hi John.

    ---
     man/Makefile.am | 14 +++++++++-----
     1 file changed, 9 insertions(+), 5 deletions(-)
    
    diff --git a/man/Makefile.am b/man/Makefile.am
    index 1627b1f..30d0319 100644
    --- a/man/Makefile.am
    +++ b/man/Makefile.am
    @@ -23,10 +23,14 @@ MAINTAINERCLEANFILES = $(man_MANS)
     # The man pages depend on the version number.
     common_mandeps = $(top_srcdir)/configure.ac
     
    -# Depend on the source, not the binary; we don't need to regenerate
    -# the binary when any source file changes, only the main one.
    +# Jose writes:
    +##  Depend on the source, not the binary; we don't need to regenerate
    +##  the binary when any source file changes, only the main one.
    +# However, there is no guarantee that the binary will exist, so dependence
    +# on the source is appropriate, and help2man is cheap anyway.

You mean "dependence on the object is appropriate"?

    +
     # Use -o so that the `missing' program can infer the output file.
     
    -poke.1: $(top_srcdir)/src/poke.c $(common_mandeps)
    -   $(HELP2MAN) -p poke --name="The GNU extensible binary editor" \
    -           $(builddir)/../src/poke -o $@
    +poke.1: $(top_builddir)/src/poke $(common_mandeps)
    + POKEDATADIR=$(abs_top_srcdir)/src $(HELP2MAN) -p poke --name="The
    GNU extensible binary editor" \
    +           $< -o $@

I don't get it.  The man/ subdirectory is recursed by Automake after
src/ while executing the target `all' so the binary should indeed exist
at the time the rule is triggered...  at least that is what happens in
my system.