Date: Mon, 24 Aug 1998 22:47:59 -0400 (EDT) From: Tim Vanderhoek <ac199@hwcn.org> To: Satoshi Asami <asami@FreeBSD.ORG> Cc: ports@FreeBSD.ORG Subject: Re: cvs commit: ports/x11-wm/afterstep/pkg PLIST (more bsd.port.mk mods) Message-ID: <Pine.BSF.3.96.980824214720.7081A-100000@localhost> In-Reply-To: <199808250146.SAA00195@silvia.hip.berkeley.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 24 Aug 1998, Satoshi Asami wrote:
> I only sent that do you because I had the feeling that we're just
> boring everyone else on the list. ;)
As opposed to the never-ending stream of PRs changes and
bugreports? ;-)
> * even I haven't tested this quite as much I would like, I'll post
> * this now in hopes of saving a testrun.
> *
> * [I'll read those latest diffs right after this... :-]
>
> I think mine are child's play compared to your stuff.
Hehe. You _did_ ask for an MLINKS variable. :)
Seriously, when I started I thought it would be just a simple
shell script tacked onto generate-plist: and man-compress:. I
hadn't intended to write some 94 lines. :(
> Do we really need the "->"? The ones in /usr/src don't have that and
> will just link even-numbered entries to odd-numbered ones. (*)
I hate reading the ones that use bsd.man.mk. This was an attempt
at an improvement. :-)
+__fl!= ${ECHO} '${MLINKS:S/ / /g}' | \
+ ${SED} -e 's/\([^ ]*\) *\-\> *\([^ ]*\)/\1/g'
+__tl!= ${ECHO} '${MLINKS:S/ / /g}' | \
+ ${SED} -e 's/\([^ ]*\) *\-\> *\([^ ]*\)/\2/g'
Those are the only two lines that require it. It _should_ work
fine if you just delete the arrows or make them optional,
although that will make an already minimal amount of error
checking even more minimal.
> Anyway, the patch did apply to my latest bsd.port.mk. The build is
> still humming along despite me doing a "make install" from
> /usr/src/share/mk, so it hasn't broken anything obvious (yet).
All my new variables do seem to slow things down a little,
though. :-(
> * Thoughts, vetoes, comments, curses? ;-)
>
> Not sure what to say. If you can send me some patches to Makefiles,
> I'll test it. (And an answer to * above too....)
There's a missing line (Sorry...). :(
What currently reads
generate-plist:
[...]
${SED} ${_sedsubplist} ${PLIST} >> ${TMPPLIST}
@for link in ${_HMLINKS}; do \
should read
generate-plist:
[...]
${SED} ${_sedsubplist} ${PLIST} >> ${TMPPLIST}
@${ECHO} @cwd / >> ${TMPPLIST}
@for link in ${_HMLINKS}; do \
That ``@cwd'' is a little evil, but it's not too easy to avoid
it. A simple hack to avoid it would just
${AppropriateVariables:S/${PREFIX}//} but...
I'll give you some diffs to make some current ports use MLINKS
for testing. How many do you want?
Okay, since as you point-out, this isn't entirely obvious to the
casual reader, I'll make some comments on how it's _supposed_ to
work... Hopefully this'll encourage a couple lurkers on -ports
to offer their opinions on whether or not this is worthwhile. :)
+# Parses space-delimited lists
+_WORD= `${ECHO} "SENT" | ${SED} -e 's/\([^ ]* *\)\{'$$((NUMB-1))'\}\([^ ]*\).*/\2/'`
+
So given a list like
list= aaa bbb ccc
One can do
@${ECHO} ${WORD:S/SENT/${list}/:S/NUMB/2/}
and print "bbb".
+.if defined(MLINKS)
+__fl!= ${ECHO} '${MLINKS:S/ / /g}' | \
+ ${SED} -e 's/\([^ ]*\) *\-\> *\([^ ]*\)/\1/g'
+__tl!= ${ECHO} '${MLINKS:S/ / /g}' | \
+ ${SED} -e 's/\([^ ]*\) *\-\> *\([^ ]*\)/\2/g'
+__fs!= ${ECHO} ${__fl} | ${SED} -e 's/[^ ]*\.\([1-9ln]\)[^. ]*/\1/g'
+__ts!= ${ECHO} ${__tl} | ${SED} -e 's/[^ ]*\.\([1-9ln]\)[^. ]*/\1/g'
Break the MLINKS variable up into
__fl -- the _list_ of each manpage on the left of a "->"
__tl -- the _list_ of each manpage on the right of a "->"
__fs -- A _list_ of man sections (1-9ln), corresponding to __fl
__ts -- A _list_ of man sections (1-9ln), corresponding to __tl
For a manpage such as qt.234.6qt, we grab the "6".
[This doesn't handle something like "iamnotamanpage" at
all---I should fix this...]
+__count:= 0
+.for __fsect in ${__fs}
+__count!= ${ECHO} $$((${__count}+1))
+__fname!= ${ECHO} ${_WORD:S/SENT/${__fl}/:S/NUMB/${__count}/}
+__tname!= ${ECHO} ${_WORD:S/SENT/${__tl}/:S/NUMB/${__count}/}
+__tsect!= ${ECHO} ${_WORD:S/SENT/${__ts}/:S/NUMB/${__count}/}
Loop once for each set of two manpages.
__fname --- grab the appropriate left-side manpage (as per __count)
__fsect --- man section ${__fname} belongs to
__tname --- grab the appropriate right-side manpage (as per __count)
__tsect --- man section ${__tname} belongs to
+.if defined(SYMMLINKS)
+__defln:= -fs
+.else
+__defln:= -f
+.endif
Default linking style. Duh. ;-)
+.else defined(NOMANCOMPRESS)
[...]
The goal here is to produce variables
_FMLINKS= /full/path/from.1 /full/path/from.2 [...]
_TMLINKS= /full/path/to.1 /full/path/to.2 [...]
_HMLINKS= -fs -fs [...]
+.if ${__tname:M/*}x == x
+.for __tsect in ${__tsect}
+_TMLINKS!= ${ECHO} ${_TMLINKS} ${MAN${__tsect:S/l/L/:S/n/N/}PREFIX}/man/${lang}/man${__tsect}/${__tname}.gz
.endfor
+.else Not an Absolute Path
+_TMLINKS!= ${ECHO} ${_TMLINKS} ${__tname}.gz
+__defln:= -fs
+.endif
This expands the short path "man.1" to its full path (as
necessary) and adds it to the appropriate list.
Don't even ask why that inner .for loop is necessary...
+.endif defined(NOMANCOMPRESS)
+_HMLINKS+= ${__defln}
+_HMLINKS:= ${_HMLINKS}
Add the link style to the appropriate list.
+.endfor __fsect in ${__fs}
+.endif defined(MLINKS)
+.endfor lang in ${MANLANG}
Loop again for the next pair of manpages.
End of conditional MLINKS section.
Loop again for next MANLANG.
Easy. :-)
+__w1!= ${ECHO} ${_HMLINKS} | ${WC} -w
+__w2!= ${ECHO} ${_FMLINKS} | ${WC} -w
+__w3!= ${ECHO} ${_TMLINKS} | ${WC} -w
+.if ${__w1} != ${__w2} || ${__w1} != ${__w3}
+.BEGIN:
+ @${ECHO_MSG} "Error: variables derived from MLINKS are bad."
+ @${FALSE}
+.endif
Make sure nobody is trying to go behind our back and cause
trouble.
Now that our variables _HMLINKS, _FMLINKS, and _TMLINKS are
setup, the rest is easy.
Pop one value off each, combine, do appropriate thing for
generate-plist: or man-compress:, repeat until finished.
How's that?
--
This .sig is not innovative, witty, or profund.
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-ports" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.96.980824214720.7081A-100000>