Skipping certain files during distfile extraction

[Roland Illig <roland.illig%gmx.de@localhost>]

Hi,

when the portability check fails for a shell program, one approach is to
patch it, the other is to ignore it.

There's a third way, which is to not extract that file at all. Currently
that would involve setting two variables:

EXTRACT_OPTS+=          -x
EXTRACT_ELEMENTS+=      contrib/*.sh

The attached patch offers a new variable called EXTRACT_SKIP that makes
such tasks simpler:

EXTRACT_SKIP=           contrib/*.sh

The latter variant is clearly more expressive. To understand the former
variant, one has to know that the option -x means to invert.

Curiously, none of the pkgsrc packages use that option at all right now.
It looks useful to me though.

One use case that I didn't explicitly mention in the documentation is to
use EXTRACT_SKIP instead of CHECK_PORTABILITY_SKIP if the file is not
needed at all. It's a somewhat special case of the "programs that are
not used".

Opinions?

Index: mk/extract/extract.mk
===================================================================
RCS file: /cvsroot/pkgsrc/mk/extract/extract.mk,v
retrieving revision 1.41
diff -u -r1.41 extract.mk
--- mk/extract/extract.mk	20 Mar 2020 15:00:45 -0000	1.41
+++ mk/extract/extract.mk	20 Mar 2020 18:00:15 -0000
@@ -32,6 +32,16 @@
 #	when using EXTRACT_CMD_DEFAULT.  By default, this is empty, which
 #	causes all files within the archive to be extracted.
 #
+#    EXTRACT_SKIP is a list of filename patterns that are skipped when
+#	extracting the distfiles. Typical use cases are:
+#
+#	* A package includes a bundled subproject that is not needed for
+#	  pkgsrc since it is already provided by a different package.
+#	* A package contains programs that are only needed for internal
+#	  project management and that are not used during the build.
+#
+#	Examples: contrib *.sh
+#
 # The following are read-only variables that may be used within a package
 #	Makefile:
 #
@@ -199,6 +209,15 @@
 _EXTRACT_TAR=
 .endif
 
+.if !empty(EXTRACT_SKIP)
+.  if empty(EXTRACT_ELEMENTS)
+EXTRACT_OPTS+=		-x
+EXTRACT_ELEMENTS=	${EXTRACT_SKIP:@pattern@${pattern:Q}@}
+.  else
+PKG_FAIL_REASON+=	"[extract.mk] EXTRACT_SKIP cannot be combined with EXTRACT_ELEMENTS."
+.  endif
+.endif
+
 .if !empty(_EXTRACT_TAR)
 EXTRACT_OPTS+=	-t ${_EXTRACT_TAR}
 .endif