Re: [lmi] Trivial patch to fix autotools builds

[Vadim Zeitlin]

On Thu, 30 May 2019 22:14:58 +0000 Greg Chicares <address@hidden> wrote:

GC> On 2019-05-29 00:23, Vadim Zeitlin wrote:
GC> > 
GC> >  I made https://github.com/vadz/lmi/pull/111 to fix the build using
GC> > autotools as currently it fails due to the recently added extra
GC> > backslashes.
GC> 
GC> Cherry-picked and pushed.

 Thanks!

GC> The proposal is:
GC> 
GC>   EMPTY =
GC>   foo_SOURCES = \
GC>      fileA.c \
GC>      fileB.c \
GC>      fileC.c \
GC>      $(EMPTY)
GC> 
GC> Probably we should prefer a name that's not a common word like
GC> "EMPTY"; is there a recommended automake convention for choosing
GC> unique names, say, "LMI_EMPTY"?
GC> 
GC> I'd be delighted to make such a change and never have to deal with
GC> this automake craziness again. What do you think?

 To be honest, I'd strongly prefer to keep Automake happy by not using
slashes on the last line than adding many useless lines consisting of just
$(EMPTY). I understand the benefits of using trailing backslashes on all
lines, but IMO they're not enough to outweigh the drawback of adding all
these lines, especially considering that many variables in Makefile.am are
defined on 3-4 lines only, so adding this dummy line and the following
blank line would inflate their line count by 50% (and for some of them it
will be 100%).

 If you do prefer to add them, the dummy variable should probably be called
something like END_OF_FILES_LIST_MARKER abbreviated to whichever you
consider to be the best balance between brevity and clarity because the
fact that a variable called EMPTY is empty can be discovered by just
locating its definition, while this doesn't help at all with understanding
_why_ is an empty variable used here. Hopefully a name similar to the one
proposed above would be more clear from this point of view.

 Regards,
VZ

pgp1ZnJU7YYPH.pgp
Description: PGP signature