Don't clean C++ submodules by default

[burg]

Accidentally cleaning mozjs or skia can cost a few minutes of your time. I will add a list of slow submodules to clean.mk, which will be skipped by make clean but not by make clean-all.

[jdm]

make cleaner

[BrendanEich]

Do we aspire to follow GNU conventions? It's ok to have local variations so long as known -- the trouble comes when someone has a stale submodule and doesn't know to use '-er' or '-all' :-/.

/be

[burg]

Is there a good summary of these conventions somewhere? I'm not familiar.

[BrendanEich]

http://www.gnu.org/prep/standards/html_node/Standard-Targets.html
ftp://ftp.gnu.org/pub/old-gnu/Manuals/make-3.79.1/html_chapter/make_14.html

Looks like mostlyclean might be the GNU-standard target to use that leaves the sub-modules intact.

/be

[burg]

What does SM, Gecko, etc do in this situation? I'm afraid that even fewer people would guess mostlyclean than cleaner or clean-all.

[BrendanEich]

SM and Gecko use autoconf213 and have clean distclean and other targets (some grandfathered and not in GNU's conventions, e.g. clobber).

My point is not what people would guess, it's what people likely expect from nearby projects with make target conventions. Guessing is not going to work well here anyway.

If we have a new circumstance that favors a longer target for "reallyclean", perhaps it is worth breaking with convention(s). There are so many to break from ;-).

/be

[burg]

For now, i've implemented clean-fast as the expedited option, while clean blows away everything.