* bug#32050: invoke (guix build utils) error reporting lacking
@ 2018-07-04 12:50 Maxim Cournoyer
[not found] ` <handler.32050.B.153070867017871.ack@debbugs.gnu.org>
0 siblings, 1 reply; 6+ messages in thread
From: Maxim Cournoyer @ 2018-07-04 12:50 UTC (permalink / raw)
To: 32050
The error reported by `invoke' isn't accurate, and mutes important
output from the program invoked.
Example of a failed tar extraction:
--8<---------------cut here---------------start------------->8---
/gnu/store/xlj36j9dw581iah51iyjgi75iikrj55h-module-import/guix/build/utils.scm:616:6: In procedure invoke:
Throw to key `srfi-34' with args `(#<condition &invoke-error [program: "tar" arguments: ("xf" "/gnu/store/yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2" "--strip-components=1" "-C" "/gnu/store/rw7yzjk1wffza65z6l6rnn25qg97jvh1-docbook-dsssl-doc-1.79" "docbook-dssl-1.79/doc") exit-status: 127 term-signal: #f stop-signal: #f] 68fc80>)'.
--8<---------------cut here---------------end--------------->8---
While reproducing the same error in an containerized environment,
calling the command directly yields:
--8<---------------cut here---------------start------------->8---
tar xf yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2 --strip-components=1 -C $PWD docbook-dssl-1.79/doc
tar: docbook-dssl-1.79/doc: Not found in archive
tar: Exiting with failure status due to previous errors
maxim@apteryx /tmp/docbook-dsssl-doc [env]# echo $?
2
--8<---------------cut here---------------end--------------->8---
Two observations:
1. invoke's reported exit status was wrong (127 rather than 2)
2. The useful error message "docbook-dssl-1.79/doc: Not found in
archive" was masked.
^ permalink raw reply [flat|nested] 6+ messages in thread[parent not found: <handler.32050.B.153070867017871.ack@debbugs.gnu.org>]
* bug#32050: invoke (guix build utils) error reporting lacking [not found] ` <handler.32050.B.153070867017871.ack@debbugs.gnu.org> @ 2018-07-04 13:12 ` Maxim Cournoyer 2018-07-04 17:45 ` Mark H Weaver 0 siblings, 1 reply; 6+ messages in thread From: Maxim Cournoyer @ 2018-07-04 13:12 UTC (permalink / raw) To: 32050 More problematic even, it seems "invoke" can't do everything that "system*" can? For example, the following package builds fine: --8<---------------cut here---------------start------------->8--- (define-public docbook-dsssl-doc (package (name "docbook-dsssl-doc") (version "1.79") (source (origin (method url-fetch) (uri (string-append "mirror://sourceforge/docbook/" name "/" version "/" name "-" version ".tar.bz2")) (sha256 (base32 "1plp5ngc96pbna4rwglp9glcadnirbm3hlcjb4gjvq1f8biic9lz")))) (build-system trivial-build-system) (arguments `(#:modules ((guix build utils)) #:builder (begin (use-modules (guix build utils)) (let ((source (assoc-ref %build-inputs "source")) (out (assoc-ref %outputs "out"))) (mkdir-p out) ;; Extract the "doc" subdirectory. (system* "tar" "xf" source "--strip-components=1" "--no-same-owner" "-C" out (string-append "docbook-dsssl-" ,version "/doc")))))) (native-inputs `(("bzip2", bzip2) ("tar" ,tar))) (home-page "https://docbook.org/") (synopsis "DocBook DSSSL style sheets documentation") (description "Documentation for the DocBook DSSSL style sheets.") (license non-copyleft))) --8<---------------cut here---------------end--------------->8--- ,run-in-store (build-package docbook-dsssl-doc #:use-substitutes? #t #:dry-run? #f) /gnu/store/ljr8xn3qbr3lahr29zc0lbki9x33zls6-docbook-dsssl-doc-1.79 $4 = #t But changing just "system*" to "invoke", I get: --8<---------------cut here---------------start------------->8--- @ build-remote /gnu/store/fdcg5nazw2jvpb7k4mq7q0750gq22zh7-docbook-dsssl-doc-1.79.drv 192.168.1.105 Backtrace: 2 (primitive-load "/gnu/store/06zszks9mwp52gih8j0mj5bahfi?") In ice-9/eval.scm: 191:35 1 (_ #f) In /gnu/store/xlj36j9dw581iah51iyjgi75iikrj55h-module-import/guix/build/utils.scm: 616:6 0 (invoke _ . _) /gnu/store/xlj36j9dw581iah51iyjgi75iikrj55h-module-import/guix/build/utils.scm:616:6: In procedure invoke: Throw to key `srfi-34' with args `(#<condition &invoke-error [program: "tar" arguments: ("xf" "/gnu/store/yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2" "--strip-components=1" "--no-same-owner" "-C" "/gnu/store/83yjsgvvapffhklg8md5cqa6gmb2dmqg-docbook-dsssl-doc-1.79" "docbook-dsssl-1.79/doc") exit-status: 127 term-signal: #f stop-signal: #f] 68fcc0>)'. builder for `/gnu/store/fdcg5nazw2jvpb7k4mq7q0750gq22zh7-docbook-dsssl-doc-1.79.drv' failed with exit code 1 derivation '/gnu/store/fdcg5nazw2jvpb7k4mq7q0750gq22zh7-docbook-dsssl-doc-1.79.drv' offloaded to '192.168.1.105' failed: build of `/gnu/store/fdcg5nazw2jvpb7k4mq7q0750gq22zh7-docbook-dsssl-doc-1.79.drv' failed While executing meta-command: Throw to key `srfi-34' with args `(#<condition &nix-protocol-error [message: "build of `/gnu/store/fdcg5nazw2jvpb7k4mq7q0750gq22zh7-docbook-dsssl-doc-1.79.drv' failed" status: 100] a7228a0>)'. --8<---------------cut here---------------end--------------->8--- Thanks, Maxim ^ permalink raw reply [flat|nested] 6+ messages in thread
* bug#32050: invoke (guix build utils) error reporting lacking 2018-07-04 13:12 ` Maxim Cournoyer @ 2018-07-04 17:45 ` Mark H Weaver 2018-07-08 15:59 ` Maxim Cournoyer 0 siblings, 1 reply; 6+ messages in thread From: Mark H Weaver @ 2018-07-04 17:45 UTC (permalink / raw) To: Maxim Cournoyer; +Cc: 32050 Hi Maxim, Maxim Cournoyer <maxim.cournoyer@gmail.com> writes: > More problematic even, it seems "invoke" can't do everything that > "system*" can? > > For example, the following package builds fine: > > (define-public docbook-dsssl-doc > (package > (name "docbook-dsssl-doc") > (version "1.79") > (source (origin > (method url-fetch) > (uri (string-append "mirror://sourceforge/docbook/" > name "/" version "/" > name "-" version ".tar.bz2")) > (sha256 > (base32 > "1plp5ngc96pbna4rwglp9glcadnirbm3hlcjb4gjvq1f8biic9lz")))) > (build-system trivial-build-system) > (arguments > `(#:modules ((guix build utils)) > #:builder > (begin > (use-modules (guix build utils)) > (let ((source (assoc-ref %build-inputs "source")) > (out (assoc-ref %outputs "out"))) > (mkdir-p out) > ;; Extract the "doc" subdirectory. > (system* "tar" "xf" source "--strip-components=1" > "--no-same-owner" "-C" out > (string-append "docbook-dsssl-" ,version "/doc")))))) > (native-inputs > `(("bzip2", bzip2) > ("tar" ,tar))) > (home-page "https://docbook.org/") > (synopsis "DocBook DSSSL style sheets documentation") > (description "Documentation for the DocBook DSSSL style sheets.") > (license non-copyleft))) > > > ,run-in-store (build-package docbook-dsssl-doc #:use-substitutes? #t #:dry-run? #f) > /gnu/store/ljr8xn3qbr3lahr29zc0lbki9x33zls6-docbook-dsssl-doc-1.79 > $4 = #t Did you check to see if the extraction actually succeeded? In this case, I don't see how it could have. When using 'tar' from within 'trivial-build-system', you need to add 'bzip2' to PATH, or else 'tar' will fail because it can't find 'bzip2'. There are many examples of setting PATH in (gnu packages fonts). Because of the way you are using 'system*' here without the usual 'zero?' wrapped around it, the numeric result code is returned as the result of the builder, and all numbers (even 0) are considered true in Scheme. So, this build will always succeed, ignoring errors from 'tar'. > But changing just "system*" to "invoke", I get: > > @ build-remote /gnu/store/fdcg5nazw2jvpb7k4mq7q0750gq22zh7-docbook-dsssl-doc-1.79.drv 192.168.1.105 > Backtrace: > 2 (primitive-load "/gnu/store/06zszks9mwp52gih8j0mj5bahfi?") > In ice-9/eval.scm: > 191:35 1 (_ #f) > In /gnu/store/xlj36j9dw581iah51iyjgi75iikrj55h-module-import/guix/build/utils.scm: > 616:6 0 (invoke _ . _) > > /gnu/store/xlj36j9dw581iah51iyjgi75iikrj55h-module-import/guix/build/utils.scm:616:6: In procedure invoke: > Throw to key `srfi-34' with args `(#<condition &invoke-error [program: "tar" arguments: ("xf" "/gnu/store/yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2" "--strip-components=1" "--no-same-owner" "-C" "/gnu/store/83yjsgvvapffhklg8md5cqa6gmb2dmqg-docbook-dsssl-doc-1.79" "docbook-dsssl-1.79/doc") exit-status: 127 term-signal: #f stop-signal: #f] 68fcc0>)'. > builder for `/gnu/store/fdcg5nazw2jvpb7k4mq7q0750gq22zh7-docbook-dsssl-doc-1.79.drv' failed with exit code 1 > derivation '/gnu/store/fdcg5nazw2jvpb7k4mq7q0750gq22zh7-docbook-dsssl-doc-1.79.drv' offloaded to '192.168.1.105' failed: build of `/gnu/store/fdcg5nazw2jvpb7k4mq7q0750gq22zh7-docbook-dsssl-doc-1.79.drv' failed > While executing meta-command: > Throw to key `srfi-34' with args `(#<condition &nix-protocol-error [message: "build of `/gnu/store/fdcg5nazw2jvpb7k4mq7q0750gq22zh7-docbook-dsssl-doc-1.79.drv' failed" status: 100] a7228a0>)'. Changing 'system*' to 'invoke' allowed you to see that something went wrong. > The error reported by `invoke' isn't accurate, and mutes important > output from the program invoked. > > Example of a failed tar extraction: > > /gnu/store/xlj36j9dw581iah51iyjgi75iikrj55h-module-import/guix/build/utils.scm:616:6: In procedure invoke: > Throw to key `srfi-34' with args `(#<condition &invoke-error [program: "tar" arguments: ("xf" "/gnu/store/yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2" "--strip-components=1" "-C" "/gnu/store/rw7yzjk1wffza65z6l6rnn25qg97jvh1-docbook-dsssl-doc-1.79" "docbook-dssl-1.79/doc") exit-status: 127 term-signal: #f stop-signal: #f] 68fc80>)'. > > > While reproducing the same error in an containerized environment, > calling the command directly yields: > > tar xf yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2 --strip-components=1 -C $PWD docbook-dssl-1.79/doc > tar: docbook-dssl-1.79/doc: Not found in archive > tar: Exiting with failure status due to previous errors > maxim@apteryx /tmp/docbook-dsssl-doc [env]# echo $? > 2 > > Two observations: > > 1. invoke's reported exit status was wrong (127 rather than 2) > 2. The useful error message "docbook-dssl-1.79/doc: Not found in > archive" was masked. I think what happened is that 'tar' actually failed in two different ways. In the first case, where the result code was 127, it failed because 'bzip2' was not found in PATH. In the second case, it failed because "docbook-dssl-1.79/doc" was not found in the archive. Or at least that's my guess. Regards, Mark ^ permalink raw reply [flat|nested] 6+ messages in thread
* bug#32050: invoke (guix build utils) error reporting lacking 2018-07-04 17:45 ` Mark H Weaver @ 2018-07-08 15:59 ` Maxim Cournoyer 2018-07-10 2:42 ` Maxim Cournoyer 0 siblings, 1 reply; 6+ messages in thread From: Maxim Cournoyer @ 2018-07-08 15:59 UTC (permalink / raw) To: Mark H Weaver; +Cc: 32050 Hi Mark, Mark H Weaver <mhw@netris.org> writes: [...] > Did you check to see if the extraction actually succeeded? > > In this case, I don't see how it could have. When using 'tar' from > within 'trivial-build-system', you need to add 'bzip2' to PATH, or else > 'tar' will fail because it can't find 'bzip2'. There are many examples > of setting PATH in (gnu packages fonts). OK. I now understand that the trivial-build-system does exactly nothing for us, so it doesn't even sets the "PATH" environment variable. This wasn't obvious to me, thank for pointing it out! > Changing 'system*' to 'invoke' allowed you to see that something went > wrong. Right, my confusion stemmed from the unset PATH variable; tar itself would return exit code 127 (command not found) in the trivial-build-system, but I couldn't reproduce directly in a containerized guix environment because there is a profile generation hook there which does set the PATH. >> The error reported by `invoke' isn't accurate, and mutes important >> output from the program invoked. >> >> Example of a failed tar extraction: >> >> /gnu/store/xlj36j9dw581iah51iyjgi75iikrj55h-module-import/guix/build/utils.scm:616:6: In procedure invoke: >> Throw to key `srfi-34' with args `(#<condition &invoke-error >> [program: "tar" arguments: ("xf" >> "/gnu/store/yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2" >> "--strip-components=1" "-C" >> "/gnu/store/rw7yzjk1wffza65z6l6rnn25qg97jvh1-docbook-dsssl-doc-1.79" >> "docbook-dssl-1.79/doc") exit-status: 127 term-signal: #f >> stop-signal: #f] 68fc80>)'. It seems that the "system" and "system*" commands of Guile as well as our "invoke" do preserve the stdout/stderr, as when testing it manually I can see the useful "sh: tar: command not found" message: --8<---------------cut here---------------start------------->8--- scheme@(guile-user)> ,use (guix build utils) scheme@(guile-user)> (system "tar xf yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2") sh: tar: command not found $1 = 32512 scheme@(guile-user)> (invoke "tar" "xf" "yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2") In execvp of tar: No such file or directory guix/build/utils.scm:616:6: In procedure invoke: Throw to key `srfi-34' with args `(#<condition &invoke-error [program: "tar" arguments: ("xf" "yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2") exit-status: 127 term-signal: #f stop-signal: #f] 1e3dc40>)'. Entering a new prompt. Type `,bt' for a backtrace or `,q' to continue. --8<---------------cut here---------------end--------------->8--- Unfortunately, when called from the builder side of Guix, this information seems to be lost: --8<---------------cut here---------------start------------->8--- ./pre-inst-env guix build --no-build-hook --check docbook-dsssl-doc @ build-started /gnu/store/4ra70nwni8mcjyd2wfx7fqvsl4fpfs5m-docbook-dsssl-doc-1.79.drv - x86_64-linux /var/log/guix/drvs/4r//a70nwni8mcjyd2wfx7fqvsl4fpfs5m-docbook-dsssl-doc-1.79.drv.bz2 Backtrace: 2 (primitive-load "/gnu/store/kj0nr8i617gjwypalfs065jllhm?") In ice-9/eval.scm: 191:35 1 (_ #f) In /gnu/store/xlj36j9dw581iah51iyjgi75iikrj55h-module-import/guix/build/utils.scm: 616:6 0 (invoke _ . _) /gnu/store/xlj36j9dw581iah51iyjgi75iikrj55h-module-import/guix/build/utils.scm:616:6: In procedure invoke: Throw to key `srfi-34' with args `(#<condition &invoke-error [program: "tar" arguments: ("-xf" "/gnu/store/yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2" "--strip-components=1" "--no-same-owner" "-C" "/gnu/store/ppf9bhzyqphmc9cvwhq8pc3yfir2qd3m-docbook-dsssl-doc-1.79" "docbook-dsssl-1.79/doc") exit-status: 127 term-signal: #f stop-signal: #f] 68fcc0>)'. builder for `/gnu/store/4ra70nwni8mcjyd2wfx7fqvsl4fpfs5m-docbook-dsssl-doc-1.79.drv' failed with exit code 1 @ build-failed /gnu/store/4ra70nwni8mcjyd2wfx7fqvsl4fpfs5m-docbook-dsssl-doc-1.79.drv - 1 builder for `/gnu/store/4ra70nwni8mcjyd2wfx7fqvsl4fpfs5m-docbook-dsssl-doc-1.79.drv' failed with exit code 1 guix build: error: build failed: build of `/gnu/store/4ra70nwni8mcjyd2wfx7fqvsl4fpfs5m-docbook-dsssl-doc-1.79.drv' failed --8<---------------cut here---------------end--------------->8--- Could it be preserved? It'd be very useful when debugging :) Maxim ^ permalink raw reply [flat|nested] 6+ messages in thread
* bug#32050: invoke (guix build utils) error reporting lacking 2018-07-08 15:59 ` Maxim Cournoyer @ 2018-07-10 2:42 ` Maxim Cournoyer 2020-05-28 4:42 ` Maxim Cournoyer 0 siblings, 1 reply; 6+ messages in thread From: Maxim Cournoyer @ 2018-07-10 2:42 UTC (permalink / raw) To: Mark H Weaver; +Cc: 32050 Maxim Cournoyer <maxim.cournoyer@gmail.com> writes: [...] > It seems that the "system" and "system*" commands of Guile as well as > our "invoke" do preserve the stdout/stderr, as when testing it manually > I can see the useful "sh: tar: command not found" message: > > scheme@(guile-user)> ,use (guix build utils) > scheme@(guile-user)> (system "tar xf yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2") > sh: tar: command not found > $1 = 32512 > scheme@(guile-user)> (invoke "tar" "xf" "yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2") > In execvp of tar: No such file or directory > guix/build/utils.scm:616:6: In procedure invoke: > Throw to key `srfi-34' with args `(#<condition &invoke-error [program: "tar" arguments: ("xf" "yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2") exit-status: 127 term-signal: #f stop-signal: #f] 1e3dc40>)'. > > Entering a new prompt. Type `,bt' for a backtrace or `,q' to continue. > > > Unfortunately, when called from the builder side of Guix, this > information seems to be lost: > > ./pre-inst-env guix build --no-build-hook --check docbook-dsssl-doc > @ build-started /gnu/store/4ra70nwni8mcjyd2wfx7fqvsl4fpfs5m-docbook-dsssl-doc-1.79.drv - x86_64-linux /var/log/guix/drvs/4r//a70nwni8mcjyd2wfx7fqvsl4fpfs5m-docbook-dsssl-doc-1.79.drv.bz2 > Backtrace: > 2 (primitive-load "/gnu/store/kj0nr8i617gjwypalfs065jllhm?") > In ice-9/eval.scm: > 191:35 1 (_ #f) > In /gnu/store/xlj36j9dw581iah51iyjgi75iikrj55h-module-import/guix/build/utils.scm: > 616:6 0 (invoke _ . _) > > /gnu/store/xlj36j9dw581iah51iyjgi75iikrj55h-module-import/guix/build/utils.scm:616:6: In procedure invoke: > Throw to key `srfi-34' with args `(#<condition &invoke-error [program: "tar" arguments: ("-xf" "/gnu/store/yanfzfi7rdyf5scf5h65xbkn7xg0nnbn-docbook-dsssl-doc-1.79.tar.bz2" "--strip-components=1" "--no-same-owner" "-C" "/gnu/store/ppf9bhzyqphmc9cvwhq8pc3yfir2qd3m-docbook-dsssl-doc-1.79" "docbook-dsssl-1.79/doc") exit-status: 127 term-signal: #f stop-signal: #f] 68fcc0>)'. > builder for `/gnu/store/4ra70nwni8mcjyd2wfx7fqvsl4fpfs5m-docbook-dsssl-doc-1.79.drv' failed with exit code 1 > @ build-failed /gnu/store/4ra70nwni8mcjyd2wfx7fqvsl4fpfs5m-docbook-dsssl-doc-1.79.drv - 1 builder for `/gnu/store/4ra70nwni8mcjyd2wfx7fqvsl4fpfs5m-docbook-dsssl-doc-1.79.drv' failed with exit code 1 > guix build: error: build failed: build of > `/gnu/store/4ra70nwni8mcjyd2wfx7fqvsl4fpfs5m-docbook-dsssl-doc-1.79.drv' > failed > > Could it be preserved? It'd be very useful when debugging :) I've investigated a bit, and as far as I can see by reading guix sources, the build log output would be displayed and any error reported by the `process-stderr' procedure in (guix store). This procedure behavior seems to depend closely on that of the guix daemon (server), so the problem could lie on the daemon side as well, so that's where I'll try digging next. Maxim ^ permalink raw reply [flat|nested] 6+ messages in thread
* bug#32050: invoke (guix build utils) error reporting lacking 2018-07-10 2:42 ` Maxim Cournoyer @ 2020-05-28 4:42 ` Maxim Cournoyer 0 siblings, 0 replies; 6+ messages in thread From: Maxim Cournoyer @ 2020-05-28 4:42 UTC (permalink / raw) To: Mark H Weaver; +Cc: 32050-done Hello, I think my original observations must have occured in Emacs while using Geiser. It has to do with what (current-output-port) is bound to in the REPL (see: https://github.com/jaor/geiser/issues/83). I'm closing, as I haven't seen this issue in almost 2 years when using Guix from the command line. Maxim ^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2020-05-28 4:43 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-07-04 12:50 bug#32050: invoke (guix build utils) error reporting lacking Maxim Cournoyer
[not found] ` <handler.32050.B.153070867017871.ack@debbugs.gnu.org>
2018-07-04 13:12 ` Maxim Cournoyer
2018-07-04 17:45 ` Mark H Weaver
2018-07-08 15:59 ` Maxim Cournoyer
2018-07-10 2:42 ` Maxim Cournoyer
2020-05-28 4:42 ` Maxim Cournoyer
Code repositories for project(s) associated with this public inbox https://git.savannah.gnu.org/cgit/guix.git This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).