These warnings are emitted during make index: sh: /usr/local/llvm19/bin/clang: not found make[4]: "/usr/ports/Mk/Uses/compiler.mk" line 86: warning: "/usr/local/llvm19/bin/clang --version" returned non-zero status make[4]: "/usr/ports/Mk/Uses/compiler.mk" line 134: warning: "/usr/local/llvm19/bin/clang++ -### /dev/null 2>&1" returned non-zero status
While checking another system, I learned these warnings are somewhere in the graphics category. --- describe.graphics --- /rescue/sh: /usr/local/llvm19/bin/clang: not found make[4]: /usr/ports/Mk/Uses/compiler.mk:86: warning: Command "/usr/local/llvm19/bin/clang --version" exited with status 127 make[4]: /usr/ports/Mk/Uses/compiler.mk:134: warning: Command "/usr/local/llvm19/bin/clang++ -### /dev/null 2>&1" exited with status 127 Done.
https://cgit.freebsd.org/ports/commit/?id=ea8d13f6c1020424c8e6bae5c9ffd8a002717275 ?
===> darktable sh: /usr/local/llvm19/bin/clang: not found make[1]: "/home/freebsd/freebsd-ports/Mk/Uses/compiler.mk" line 86: warning: "/usr/local/llvm19/bin/clang --version" returned non-zero status make[1]: "/home/freebsd/freebsd-ports/Mk/Uses/compiler.mk" line 134: warning: "/usr/local/llvm19/bin/clang++ -### /dev/null 2>&1" returned non-zero status
I still get these warnings. Any updates on this issue?
(In reply to Wolfram Schneider from comment #4) The following patch silences portlint in graphics/darktable and 'make index': diff --git a/graphics/darktable/Makefile b/graphics/darktable/Makefile index 1dc5eb14e03f..c8d11bc7d96a 100644 --- a/graphics/darktable/Makefile +++ b/graphics/darktable/Makefile @@ -104,6 +104,8 @@ PLIST_SUB+= NO_OPENCL="@comment " PLIST_SUB+= NO_OPENCL="" .endif +.include <bsd.port.pre.mk> + .if ${ARCH} == powerpc64le LDFLAGS+= -lexecinfo USES+= compiler:gcc-c++11-lib @@ -111,8 +113,6 @@ USES+= compiler:gcc-c++11-lib USES+= compiler:c++11-lib .endif -.include <bsd.port.pre.mk> - # work around CMAKE/CCACHE flaw (cmake stomps over ccache) .if "${WITH_CCACHE_BUILD}" == "yes" && !defined(NO_CCACHE) CMAKE_ARGS+= -DCMAKE_C_COMPILER_LAUNCHER=${CCACHE_BIN} \ The following still works: $ env ARCH=powerpc64le make -VUSES cmake desktop-file-utils gl gnome jpeg llvm:build lua perl5 pkgconfig sdl shebangfix sqlite tar:xz xorg gettext compiler:gcc-c++11-lib $ make -VUSES cmake desktop-file-utils gl gnome jpeg llvm:build lua perl5 pkgconfig sdl shebangfix sqlite tar:xz xorg gettext compiler:c++11-lib
But why is the warning only displayed on some systems? And 'make V=1 index" shows even more weird messages: make_index: ipxe-g20250401,1: no entry for /usr/ports/lang/gcc1 make_index: ipxe-g20250401,1: no entry for /usr/ports/lang/gcc1 make_index: gretl-2024d_2: no entry for /usr/ports/lang/gcc1 make_index: gretl-2024d_2: no entry for /usr/ports/lang/gcc1 make_index: tigcc-0.96.b8_11: no entry for /usr/ports/lang/gcc1 make_index: tigcc-0.96.b8_11: no entry for /usr/ports/lang/gcc1 [...] No idea where it gets gcc1 from.
(In reply to Herbert J. Skuhra from comment #6) Sorry, please ignore the 2nd part of my previous message (no entry for gcc1). :-(
I can’t reproduce the problem with `make index` on my local Ports tree copy. Is it with some specific architecture? Any step I am missing?
(In reply to Jean-Sébastien Pédron from comment #8) I get this problem on all supported FreeBSD releases (stable 13/14/15 current) on intel and arm if llvm19 is *not* installed on the local machine. $ pkg info llvm19 pkg: No package(s) matching llvm19 $ /usr/local/llvm19/bin/clang sh: /usr/local/llvm19/bin/clang: not found
Has anyone tried the patch in comment #5? Is it wrong? Yes, this a general issue (main, stable15, amd64/aarch64) I didn't get error on some systems because a different version of llvm was installed and DEFAULT_VERSION+= llvm=XX was used.
*** Bug 292505 has been marked as a duplicate of this bug. ***
(In reply to Wolfram Schneider from comment #9) If this is the same kind of problem I used to fight, the clue is to understand where in bsd.*.mk variables are defined. If you use variables that are only defined in bsd.port.pre.mk and you haven't included that file yet, wierd things can happen. And 'make -V' will lie to you, because it will give you the values the variables have after the entire Makefile has run, including bsd.port.post.mk. So careful examination of bsd.port.pre.mk and whatever it brings in, is recommended.
(In reply to Herbert J. Skuhra from comment #10) The patch looks good, moving .include <bsd.port.pre.mk> some lines up in graphics/darktable/Makefile solves the problem for me.
I didn’t try your patch, Herbert J. Skuhra, yet, as I wanted to find something more specific, based on Mark Linimon’s comment. Here is what I came up with: ---8<--- diff --git a/graphics/darktable/Makefile b/graphics/darktable/Makefile index 1dc5eb14e03f..0518786fa1b5 100644 --- a/graphics/darktable/Makefile +++ b/graphics/darktable/Makefile @@ -42,8 +42,10 @@ LIB_DEPENDS= libavif.so:graphics/libavif \ libtiff.so:graphics/tiff \ libwebp.so:graphics/webp -USES= cmake desktop-file-utils gl gnome jpeg llvm:build lua perl5 \ - pkgconfig sdl shebangfix sqlite tar:xz xorg +# Note that compiler-related Uses are defined later in this Makefile, based on +# the value of `${ARCH}`. +USES= cmake desktop-file-utils gl gnome jpeg lua perl5 pkgconfig sdl \ + shebangfix sqlite tar:xz xorg USE_GL= gl USE_GNOME= cairo gtk30 intltool librsvg2 libxml2 USE_LDCONFIG= ${PREFIX}/lib/${PORTNAME} @@ -110,6 +112,12 @@ USES+= compiler:gcc-c++11-lib .else USES+= compiler:c++11-lib .endif +# `llvm:build` must be added after `compiler:*` above. Otherwise it triggers +# the following warning from Mk/compiler.mk: +# +# make: .../Mk/Uses/compiler.mk:86: warning: Command "/usr/local/llvm19/bin/clang --version" exited with status 127 +# make: .../Mk/Uses/compiler.mk:134: warning: Command "/usr/local/llvm19/bin/clang++ -### /dev/null 2>&1" exited with status 127 +USES+= llvm:build .include <bsd.port.pre.mk> ---8<--- Apparently, the order in USES= is significant for the compiler-related Uses. If `llvm:build` is added after `compiler:*`, the warning is gone. I’m building the port in Poudriere to confirm. The initial order was based on alphabetical order, so `compiler:*` was before `llvm:*` by chance. After the fix in https://cgit.freebsd.org/ports/commit/?id=ea8d13f6c1020424c8e6bae5c9ffd8a002717275, `compiler:*` was added after.
(In reply to Jean-Sébastien Pédron from comment #14) I've just verified placing USES=llvm_build after USES=compiler:* for graphics/darktable silences the error messages I've seen lately. diff --git a/graphics/darktable/Makefile b/graphics/darktable/Makefile index 1dc5eb14e03f..04c18dc94e3d 100644 --- a/graphics/darktable/Makefile +++ b/graphics/darktable/Makefile @@ -42,7 +42,7 @@ LIB_DEPENDS= libavif.so:graphics/libavif \ libtiff.so:graphics/tiff \ libwebp.so:graphics/webp -USES= cmake desktop-file-utils gl gnome jpeg llvm:build lua perl5 \ +USES= cmake desktop-file-utils gl gnome jpeg lua perl5 \ pkgconfig sdl shebangfix sqlite tar:xz xorg USE_GL= gl USE_GNOME= cairo gtk30 intltool librsvg2 libxml2 @@ -110,6 +110,7 @@ USES+= compiler:gcc-c++11-lib .else USES+= compiler:c++11-lib .endif +USES+= llvm:build .include <bsd.port.pre.mk> root@freebsd-main-zfs:~ # make -C /usr/ports index ; echo ; date Generating INDEX-16 - please wait ... --- describe.accessibility --- --- describe.arabic --- --- describe.archivers --- --- describe.astro --- --- describe.audio --- --- describe.benchmarks --- --- describe.biology --- --- describe.cad --- --- describe.chinese --- --- describe.comms --- --- describe.converters --- --- describe.databases --- --- describe.deskutils --- --- describe.devel --- --- describe.dns --- --- describe.editors --- --- describe.emulators --- --- describe.finance --- --- describe.french --- --- describe.filesystems --- --- describe.ftp --- --- describe.games --- --- describe.german --- --- describe.graphics --- --- describe.hebrew --- --- describe.hungarian --- --- describe.irc --- --- describe.japanese --- --- describe.java --- --- describe.korean --- --- describe.lang --- --- describe.mail --- --- describe.math --- --- describe.misc --- --- describe.multimedia --- --- describe.net --- --- describe.net-im --- --- describe.net-mgmt --- --- describe.net-p2p --- --- describe.news --- --- describe.polish --- --- describe.ports-mgmt --- --- describe.portuguese --- --- describe.print --- --- describe.russian --- --- describe.science --- --- describe.security --- --- describe.shells --- --- describe.sysutils --- --- describe.textproc --- --- describe.ukrainian --- --- describe.vietnamese --- --- describe.www --- --- describe.x11 --- --- describe.x11-clocks --- --- describe.x11-drivers --- --- describe.x11-fm --- --- describe.x11-fonts --- --- describe.x11-servers --- --- describe.x11-themes --- --- describe.x11-toolkits --- --- describe.x11-wm --- --- describe.local --- Done. Wed Jan 21 10:53:51 CET 2026 root@freebsd-main-zfs:~ #
Stupid question: Is llvm:build really needed when ARCH == powerpc64le or can it be added to the line "USES+= compiler:c++11-lib" as well?
(In reply to Herbert J. Skuhra from comment #16) The relevant snippet of the Makefile looks like this: .if ${ARCH} == powerpc64le LDFLAGS+= -lexecinfo USES+= compiler:gcc-c++11-lib .else USES+= compiler:c++11-lib .endif USES+= llvm:build While we could add llvm:build after compiler:* on lines 3 and 5 above, it's more functionally correct to add it in the end. Or, maybe I misunderstood your question.
A commit in branch main references this bug: URL: https://cgit.FreeBSD.org/ports/commit/?id=50e6dae0b7b2bbf3c62ed7fe7bf5df76bd128391 commit 50e6dae0b7b2bbf3c62ed7fe7bf5df76bd128391 Author: Jean-Sébastien Pédron <dumbbell@FreeBSD.org> AuthorDate: 2026-01-22 21:07:57 +0000 Commit: Jean-Sébastien Pédron <dumbbell@FreeBSD.org> CommitDate: 2026-01-22 21:07:57 +0000 graphics/darktable: Add `llvm:build` after `compiler:*` in USES= The move of `USES+=compiler:*` later in the file to be able to check `${ARCH}` apparently triggered the following warning: make: .../Mk/Uses/compiler.mk:86: warning: Command "/usr/local/llvm19/bin/clang --version" exited with status 127 make: .../Mk/Uses/compiler.mk:134: warning: Command "/usr/local/llvm19/bin/clang++ -### /dev/null 2>&1" exited with status 127 It looks like the order of `compiler:*` and `llvm:*` is important. Moving the addition of `llvm:compiler` after `compiler:*` was added conditionally fixes that warning. PR: 292128 Reported by: Trond Endrestøl <Trond.Endrestol@ximalas.info> Tested by: Trond Endrestøl <Trond.Endrestol@ximalas.info>, Herbert J. Skuhra <herbert@gojira.at> graphics/darktable/Makefile | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-)
I committed the patch. I couldn't reproduce the problem with it. Thank you for the initial report, patches and tests!
(In reply to Trond Endrestøl from comment #17) The question was if the following would work too: .if ${ARCH} == powerpc64le LDFLAGS+= -lexecinfo USES+= compiler:gcc-c++11-lib .else USES+= compiler:c++11-lib llvm:build .endif Doesn't gcc-c++11-lib imply USE_GCC? Why is llvm from port needed on powerpc64le too?
(In reply to Herbert J. Skuhra from comment #20) Your change has shown no adverse effects on amd64. Can someone running powerpc64le verify if that platform need llvm alongside gcc when building this port?
Created attachment 267411 [details] build log without USES=llvm (In reply to Trond Endrestøl from comment #21) LLVM is not needed. I attached a build log with: diff --git a/graphics/darktable/Makefile b/graphics/darktable/Makefile index 0518786fa1b5..2c188c0a5f28 100644 --- a/graphics/darktable/Makefile +++ b/graphics/darktable/Makefile @@ -111,13 +111,13 @@ LDFLAGS+= -lexecinfo USES+= compiler:gcc-c++11-lib .else USES+= compiler:c++11-lib -.endif # `llvm:build` must be added after `compiler:*` above. Otherwise it triggers # the following warning from Mk/compiler.mk: # # make: .../Mk/Uses/compiler.mk:86: warning: Command "/usr/local/llvm19/bin/clang --version" exited with status 127 # make: .../Mk/Uses/compiler.mk:134: warning: Command "/usr/local/llvm19/bin/clang++ -### /dev/null 2>&1" exited with status 127 USES+= llvm:build +.endif .include <bsd.port.pre.mk>
(In reply to Piotr Kubaj from comment #22) Excellent. Maybe the committer would be so kind to commit the proposal found in comment 20.