Steve Hay wrote:
>
> Running "[dn]make" then "[dn]make distclean" on Win32 still doesn't
> quite put you back where you started, which could affect Test::Smoke.
>
> Various .pm files that were either generated by _pm.PL files or else
> copied from somewhere get left behind (including a whole Encode/
> directory), and a few ext/ modules leave const-c.inc and const-xs.inc
> files behind.
>
> The attached patch deletes such .pm files, and changes the Makefile.PL's
> in a few ext/ modules to remove those *.inc files when doing a "clean"
> (which "distclean" does) rather than only when doing a "realclean". (Do
> the affected modules' $VERSION's need bumping up because of these changes?)
>
> I've also removed the explicit deletion of files in directories that are
> about to be removed with "rmdir /s /q" since the /s causes the entire
> directory tree (i.e. including all files and sub-directories within in)
> to be deleted anyway.
>
> Finally, I've tidied up the end of the "distclean" target in Makefile to
> follow the rather nicer "cd DIR && del FILES" idiom employed by
> makefile.mk, rather than doing separate cd's and del's.
There are lots of chunks in your patch that go like
- realclean => {FILES=> 'const-c.inc const-xs.inc'},
+ clean => {FILES=> 'const-c.inc const-xs.inc'},
but "make distclean" is supposed to do a realclean in the extensions.
As the MakeMaker doc states,
make clean # delete local temp files (Makefile gets renamed)
make realclean # delete derived files (including ./blib)
If it doesn't, or if it does and fails to delete the files listed under
"clean", that's a MakeMaker Win32 bug. IMO.
Besides this, I applied the win32/?akefile* part of your patch as
change #22477, thanks.
