Maintainer mode

Maintainer mode

[Ben Elliston]

After thinking about it a bit, I am ready to propose that the opcodes
and sim Makefile.am's be modified so that they no longer rely on
`--enable-cgen-maint', but instead use the @MAINTAINER_MODE@
substitutions provided by AM_MAINTAINER_MODE to determine whether or
not to include rules for automatically regenerating files.

The advantages here are obvious: you need not remember an extra option
to `configure' and it works more obviously.

Any thoughts before I put it to the gdb and binutils mailing lists?
Hopefully this is a step in the right direction of eliminating the
generated files from the tree altogether.

Ben

Maintainer mode

[Doug Evans]

Ben Elliston writes:
 > After thinking about it a bit, I am ready to propose that the opcodes
 > and sim Makefile.am's be modified so that they no longer rely on
 > `--enable-cgen-maint', but instead use the @MAINTAINER_MODE@
 > substitutions provided by AM_MAINTAINER_MODE to determine whether or
 > not to include rules for automatically regenerating files.
 > 
 > The advantages here are obvious: you need not remember an extra option
 > to `configure' and it works more obviously.

I recognize the reasons why one would want to do this.
I didn't do this because it carries extra burdens I wasn't prepared
to dump on binutils/gdb people.

Maintainer mode

[Doug Evans]

Doug Evans writes:
 > Ben Elliston writes:
 >  > After thinking about it a bit, I am ready to propose that the opcodes
 >  > and sim Makefile.am's be modified so that they no longer rely on
 >  > `--enable-cgen-maint', but instead use the @MAINTAINER_MODE@
 >  > substitutions provided by AM_MAINTAINER_MODE to determine whether or
 >  > not to include rules for automatically regenerating files.
 >  > 
 >  > The advantages here are obvious: you need not remember an extra option
 >  > to `configure' and it works more obviously.
 > 
 > I recognize the reasons why one would want to do this.
 > I didn't do this because it carries extra burdens I wasn't prepared
 > to dump on binutils/gdb people.

Plus, turning on maintainer-mode often caused lots of trivial diffs
in lots of Makefile.in's/configure.in's (for me anyway).
When I'm hacking on cgen stuff, I'd rather just not be put the
the trouble of having to deal with all of that.  ymmv of course.

Re: Maintainer mode

[Ben Elliston]

Frank Ch. Eigler writes:

 > > [make --enable-maintainer-mode => --enable-cgen-maint] [...]
 > > Any thoughts before I put it to the gdb and binutils mailing lists?

 > Right - there should be no problem.

 > > Hopefully this is a step in the right direction of eliminating the
 > > generated files from the tree altogether.

 > I don't think this is a good idea.  The generated files should remain
 > in the tree for self-contained-ness, faster building, and better change
 > isolation.  (If you would like a way of regenerating all the dependent
 > files in all the source trees, then make a shell script to do
 > that.)

Now that I've had time to think about this one, the idea of
eliminating --enable-cgen-maint and removing generated files are
independent.  I am happy with your suggestion to leave the generated
files there--this is what happens now with `configure.in' and
`configure', for instance.

Ben

Re: Maintainer mode

[Ben Elliston]

Doug Evans writes:

 >  > The advantages here are obvious: you need not remember an extra option
 >  > to `configure' and it works more obviously.

 > I recognize the reasons why one would want to do this.
 > I didn't do this because it carries extra burdens I wasn't prepared
 > to dump on binutils/gdb people.

This burden only exists when the modification times of the .cpu files
is newer than the generated files.  The same burden could be said for
people building GCC and requiring Bison because a source distribution
has generated C files that are older than the `.y' file.  If it is
carefully managed, I don't think this will affect anyone.  Do you
agree?

Ben

Re: Maintainer mode

[Eric Christopher]

> 
> Now that I've had time to think about this one, the idea of
> eliminating --enable-cgen-maint and removing generated files are
> independent.  I am happy with your suggestion to leave the generated
> files there--this is what happens now with `configure.in' and
> `configure', for instance.
> 

What we did in gcc is remove some of the generated files from cvs, but
leave them in the snapshots.  The idea is that someone working on the
compiler (cvs) should have all of the tools to work on the compiler, but
that the generated files should be there for the casual user.

-eric

Re: Maintainer mode

[Doug Evans]

Ben Elliston writes:
 > This burden only exists when the modification times of the .cpu files
 > is newer than the generated files.  The same burden could be said for
 > people building GCC and requiring Bison because a source distribution
 > has generated C files that are older than the `.y' file.  If it is
 > carefully managed, I don't think this will affect anyone.  Do you
 > agree?

Try it and find out.