20.10.2024 13:58, Klemens Nanni пишет:
> 06.10.2024 23:52, Klemens Nanni пишет:
>> 'checksum' is noisy, especially when working on ports with lots of distfiles.
>> Take net/go-ipfs for example with 1996(!) files in distinfo:
>>
>> $ cat checksum.sh
>> make checksum "$@" | awk '
>> /up to date/ {up++; if (up > 1) next}
>> /SHA256.*OK/ {ok++; if (ok > 1) next}
>> {print}
>> END {print "skipped up/ok", up, ok}
>> '
>>
>> $ sh ./checksum.sh CHECKSUM_QUIET=No
>> ===> Checking files for kubo-0.26.0
>> `/home/distfiles/kubo-v0.26.0.zip' is up to date.
>> >> (SHA256) kubo-v0.26.0.zip: OK
>> skipped up/ok 1996 1996
>>
>> In this usual case, I'd like to see all good cases folded so the output fits
>> on a page, even for monster ports.
>
>> So far so good, I haven't looked into the 'up to date' part, yet.
>
> checksum is quiet now, but fetch is not:
>
> $ make fetch | grep -c 'is up to date'
> 1996
>
> These lines are produced by make(1) itself; minimal reproducer (works
> anywhere,
> needs no Makefile):
>
> $ echo 'int main(){return 0;}' >| foo.c
> $ make foo
> cc -O2 -pipe -o foo foo.c
> $ make foo
> `foo' is up to date.
>
> Here's a WIP bsd.port.mk diff filtering those lines, such that, finally, even
> huge
> ports produce a sane amount of output:
>
> $ make clean=all ; make clean=dist
> ...
> $ make fetch
> ...
> $ make checksum # runs 'fetch' internally, but now all files are there
> ===> Checking files for kubo-0.26.0
> >> (SHA256) all files: OK
>
>
> The "drawback" is that ftp(1) progress meters, if enabled, do not load
> progressively,
> since they now run through a pipe and not directly on the TTY anymore.
> Use 'make _FETCH_FILTER=' with this diff to see old/current behaviour.
>
> One fix would be to teach make(1) how to shut up on such messages.
> GNU make(1) can do that already:
>
> -s, --silent, --quiet
> Silent operation; do not print the commands as they are executed.
>
> $ gmake foo
> gmake: 'foo' is up to date.
> $ gmake foo -s ; echo $?
> 0
>
> What do you think about quieting down 'fetch' and 'fetch-all' in general?
>
> Personally, I'd be happy with the simple .mk diff below and not touching
> make(1),
> but teaching make(1) a new trick and silencing our targets properly is fun,
> too.
New version only silencing 'is up to date' under PROGRESS_METER=No, i.e. when
there's
no progressive progress bar in the first place that would turn into a simple
line due
to use of grep/no TTY.
So this includes the diff I just sent in the other thread.
Thoughts?
Index: bsd.port.mk
===================================================================
RCS file: /cvs/ports/infrastructure/mk/bsd.port.mk,v
diff -u -p -r1.1640 bsd.port.mk
--- bsd.port.mk 18 Oct 2024 22:39:30 -0000 1.1640
+++ bsd.port.mk 1 Nov 2024 19:49:08 -0000
@@ -163,12 +163,17 @@ PKG_DBDIR ?= /var/db/pkg
PROGRESS_METER ?= Yes
.if ${PROGRESS_METER:L} == "yes"
-_PROGRESS = -m
+_PKG_PROGRESS = -m
+_FETCH_PROGRESS = -m
+.elif ${PROGRESS_METER:L} == "no"
+_PKG_PROGRESS =
+_FETCH_PROGRESS = -M
.else
-_PROGRESS =
+_PKG_PROGRESS =
+_FETCH_PROGRESS =
.endif
-FETCH_CMD ?= /usr/bin/ftp -V ${_PROGRESS} -C
+FETCH_CMD ?= /usr/bin/ftp -V ${_FETCH_PROGRESS} -C
# switch for fetching each distfile: avoid warnings for missing
# distinfo and wrong size when running makesum
@@ -181,13 +186,13 @@ PKG_INFO ?= /usr/sbin/pkg_info
PKG_CREATE ?= /usr/sbin/pkg_create
PKG_DELETE ?= /usr/sbin/pkg_delete
-_PKG_ADD = ${PKG_ADD} ${_PROGRESS} -I
-_PKG_CREATE = ${PKG_CREATE} ${_PROGRESS}
+_PKG_ADD = ${PKG_ADD} ${_PKG_PROGRESS} -I
+_PKG_CREATE = ${PKG_CREATE} ${_PKG_PROGRESS}
_SUDO_PKG_ADD_LOCAL = TRUSTED_PKG_PATH=${_PKG_REPO} ${SUDO} ${_PKG_ADD}
.if ${INSTALL_DEBUG_PACKAGES:L} == "yes"
_SUDO_PKG_ADD_LOCAL += -d
.endif
-_PKG_DELETE = ${PKG_DELETE} ${_PROGRESS}
+_PKG_DELETE = ${PKG_DELETE} ${_PKG_PROGRESS}
.if defined(PKG_PATH)
_PKG_PATH = ${PKG_PATH}
@@ -2525,12 +2530,27 @@ ${_DEP${_m}WANTLIB_COOKIE}: ${_DEPBUILD_
.endfor
+# make(1) has no switch to silence general target status messages,
+# so the only way to silence them is a pipe/filter, but that means
+# FETCH_CMD aka. ftp(1) run by recursive MAKE would have no TTY,
+# i.e. its progress meter wouldn't be progressive.
+#
+# gmake(1) -s, --silent, --quiet does skip "... is up to date." messages;
+# perhaps adopt this behaviour and replace such pipe with just -s?
+#
+# Filter only if the user already disabled the only thing we'd "break".
+.if ${PROGRESS_METER:L} == "no"
+_FETCH_FILTER = | { grep -v ' is up to date\.$$' || true; }
+.else
+_FETCH_FILTER =
+.endif
+
_internal-fetch-all:
# See ports/infrastructure/templates/Makefile.template
@${ECHO_MSG} "===> Checking files for ${FULLPKGNAME}${_MASTER}"
# What FETCH-ALL normally does:
. if !empty(MAKESUMFILES)
- @${_MAKE} ${MAKESUMFILES:S@^@${DISTDIR}/@}
+ @${_MAKE} ${MAKESUMFILES:S@^@${DISTDIR}/@} ${_FETCH_FILTER}
. endif
# End of FETCH
@@ -2581,7 +2601,7 @@ _internal-fetch:
@${ECHO_MSG} "===> Checking files for ${FULLPKGNAME}${_MASTER}"
# What FETCH normally does:
. if !empty(CHECKSUMFILES)
- @${_MAKE} ${CHECKSUMFILES:S@^@${DISTDIR}/@}
+ @${_MAKE} ${CHECKSUMFILES:S@^@${DISTDIR}/@} ${_FETCH_FILTER}
. endif
# End of FETCH
