Re: Mysterious error while refactoring guix/scripts/system.scm
Hello, Chris Marusich skribis: > The reason I wanted to perform this refactoring in the first place is > because I'd like to add a new procedure to guix/scripts/system.scm > called 'switch-to-system-generation'. Because this new procedure calls > other procedures which seem to require access to the store, I thought I > would need to call 'switch-to-system-generation' via 'run-with-store'. > Am I just confused? Essentially, to write a procedure that needs to access the store, you need to write a “monadic procedure”—i.e., a procedure that returns a value in the “store monad” instead of a “normal value.” To do that, you would write: (define (switch-to-system-generation …) (with-monad %store-monad (return 'some-value)));return this symbol as a monadic value or: (define (switch-to-system-generation …) (some-monadic-procedure x y z)) ;tail-call a monadic procedure The caller of a monadic procedure must be prepared to “unpack” its value, using the monadic “bind” operator. The ‘mlet’ form does exactly that: (define (the-caller …) (mlet %store-monad ((result (switch-to-system-generation …))) …)) At the bottom, there must be somewhere a call to ‘run-with-store’ to “run” the monadic value in the monad (info "(guix) The Store Monad"). This call is already in ‘process-action’ in (guix scripts system). > In particular, 'switch-to-system-generation' will eventually call the > existing procedure 'grub-configuration-file' (defined in > gnu/system/grub.scm). As I understand it, 'grub-configuration-file' > returns a derivation that builds a GRUB configuration file. This > existing 'grub-configuration-file' procedure does a lot with the store > and gexps. I thought that if I didn't use 'run-with-store' to run > 'switch-to-system-generation', it wouldn't work because > 'grub-configuration-file' wouldn't work. ‘grub-configuration-file’ is a monadic procedure. Thus, to “unpack” its return value, you need to bind it, for instance with ‘mlet’: (mlet %store-monad ((file (grub-configuration-file …))) …) or: (with-monad %store-monad (>>= (grub-configuration-file …) (lambda (file) …))) HTH! Ludo’.
Re: Mysterious error while refactoring guix/scripts/system.scm
Hi Ludo’, Thank you for getting back to me! l...@gnu.org (Ludovic Courtès) writes: > Hi Chris, and sorry for the delay! > > Chris Marusich skribis: > >> Backtrace: >> In ice-9/boot-9.scm: >> 157: 15 [catch #t # ...] >> In unknown file: >>?: 14 [apply-smob/1 #] >> In ice-9/boot-9.scm: >> 63: 13 [call-with-prompt prompt0 ...] >> In ice-9/eval.scm: >> 432: 12 [eval # #] >> In ice-9/boot-9.scm: >> 2401: 11 [save-module-excursion #> ice-9/boot-9.scm:4045:3 ()>] >> 4050: 10 [#] >> 1724: 9 [%start-stack load-stack ...] >> 1729: 8 [#] >> In unknown file: >>?: 7 [primitive-load "/root/guix/scripts/guix"] >> In guix/ui.scm: >> 1209: 6 [run-guix-command system "list-generations"] >> In ice-9/boot-9.scm: >> 157: 5 [catch srfi-34 # ...] >> 157: 4 [catch system-error ...] >> In guix/scripts/system.scm: >> 884: 3 [#] >> 818: 2 [process-command list-generations () ...] >> In guix/store.scm: >> 1182: 1 [run-with-store # ...] >> In unknown file: >>?: 0 [# #] >> >> ERROR: In procedure #: >> ERROR: Wrong type to apply: # > > This means that we’re trying to invoke #, but > # is not a procedure. > > Note that this is from within ‘run-to-store’, which is the procedure to > “run” a monadic value. So in effect, what happened is equivalent to: > > > scheme@(guile-user)> ,use(guix monads) > scheme@(guile-user)> ,use(guix store) > scheme@(guile-user)> (with-store s > (run-with-store s *unspecified*)) > ERROR: In procedure #: > ERROR: Wrong type to apply: # > > > The bug here is in fact a type error: ‘run-with-store’ expect a monadic > value, but what we have here is a regular value. > > To “fix” it, we need: > > > scheme@(guile-user)> (with-store s > (run-with-store s >(with-monad %store-monad (return *unspecified* > > > … where ‘return’ procedure a monadic value from a normal value. > > In your case, ‘list-generations’ is not a monadic procedure (a procedure > that returns a monadic value), so in this patch: > > (run-with-store store > … > (list-generations)) > > … triggers this very type error. I see. I've read the sections in the manual about gexps, and I've peeked at the gexp code. I thought that basically I could use run-with-store to run any Guile code I want, but I seem to be missing something. I will try experimenting with monads and the store to learn more and hopefully avoid surprises like this going forward. > I imagine this may be more than you wanted to learn. ;-) > Monads in a dynamically typed setting are kinda annoying because of > this. I appreciate the info. Hopefully I will fully understand it soon. >> (define (process-command command args opts) >>"Process COMMAND, one of the 'guix system' sub-commands. ARGS is its >> argument list and OPTS is the option alist." >> - (case command >> -((list-generations) >> - ;; List generations. No need to connect to the daemon, etc. >> - (let ((pattern (match args >> - (() "") >> - ((pattern) pattern) >> - (x (leave (_ "wrong number of arguments~%")) >> - (list-generations pattern))) >> -(else >> - (process-action command args opts >> + (with-store store >> +(set-build-options-from-command-line store opts) >> + >> +(run-with-store store >> + (mbegin %store-monad >> +(set-guile-for-build (default-guile)) >> +(case command >> + ((list-generations) >> + (let ((pattern (match args >> +(() "") >> +((pattern) pattern) >> +(x (leave (_ "wrong number of arguments~%")) >> + (list-generations pattern))) >> + (else >> + (process-action command args opts >> + #:system (assoc-ref opts 'system > > As the comment above suggests, the idea here was to avoid connecting to > the daemon for operations that do not need it, such as > ‘list-generations’. I think we should preserve this property. The reason I wanted to perform this refactoring in the first place is because I'd like to add a new procedure to guix/scripts/system.scm called 'switch-to-system-generation'. Because this new procedure calls other procedures which seem to require access to the store, I thought I would need to call 'switch-to-system-generation' via 'run-with-store'. Am I just confused? In particular, 'switch-to-system-generation' will eventually call the existing procedure 'grub-configuration-file' (defined in gnu/system/grub.scm). As I understand it, 'grub-configuration-file' returns a derivation that builds a GRUB configuration file. This existing 'grub-configuration-file' procedure does a lot with the store and gexps. I thought that if I didn't use 'run-with-store' to run 'switch-to-system-generation', it wouldn't work because 'grub-configuration-file' wouldn't work. > Thanks for your work! I’m sorry this is more painful th
Re: Mysterious error while refactoring guix/scripts/system.scm
Hi Chris, and sorry for the delay! Chris Marusich skribis: > Backtrace: > In ice-9/boot-9.scm: > 157: 15 [catch #t # ...] > In unknown file: >?: 14 [apply-smob/1 #] > In ice-9/boot-9.scm: > 63: 13 [call-with-prompt prompt0 ...] > In ice-9/eval.scm: > 432: 12 [eval # #] > In ice-9/boot-9.scm: > 2401: 11 [save-module-excursion # ice-9/boot-9.scm:4045:3 ()>] > 4050: 10 [#] > 1724: 9 [%start-stack load-stack ...] > 1729: 8 [#] > In unknown file: >?: 7 [primitive-load "/root/guix/scripts/guix"] > In guix/ui.scm: > 1209: 6 [run-guix-command system "list-generations"] > In ice-9/boot-9.scm: > 157: 5 [catch srfi-34 # ...] > 157: 4 [catch system-error ...] > In guix/scripts/system.scm: > 884: 3 [#] > 818: 2 [process-command list-generations () ...] > In guix/store.scm: > 1182: 1 [run-with-store # ...] > In unknown file: >?: 0 [# #] > > ERROR: In procedure #: > ERROR: Wrong type to apply: # This means that we’re trying to invoke #, but # is not a procedure. Note that this is from within ‘run-to-store’, which is the procedure to “run” a monadic value. So in effect, what happened is equivalent to: --8<---cut here---start->8--- scheme@(guile-user)> ,use(guix monads) scheme@(guile-user)> ,use(guix store) scheme@(guile-user)> (with-store s (run-with-store s *unspecified*)) ERROR: In procedure #: ERROR: Wrong type to apply: # --8<---cut here---end--->8--- The bug here is in fact a type error: ‘run-with-store’ expect a monadic value, but what we have here is a regular value. To “fix” it, we need: --8<---cut here---start->8--- scheme@(guile-user)> (with-store s (run-with-store s (with-monad %store-monad (return *unspecified* --8<---cut here---end--->8--- … where ‘return’ procedure a monadic value from a normal value. In your case, ‘list-generations’ is not a monadic procedure (a procedure that returns a monadic value), so in this patch: (run-with-store store … (list-generations)) … triggers this very type error. I imagine this may be more than you wanted to learn. ;-) Monads in a dynamically typed setting are kinda annoying because of this. > (define (process-command command args opts) >"Process COMMAND, one of the 'guix system' sub-commands. ARGS is its > argument list and OPTS is the option alist." > - (case command > -((list-generations) > - ;; List generations. No need to connect to the daemon, etc. > - (let ((pattern (match args > - (() "") > - ((pattern) pattern) > - (x (leave (_ "wrong number of arguments~%")) > - (list-generations pattern))) > -(else > - (process-action command args opts > + (with-store store > +(set-build-options-from-command-line store opts) > + > +(run-with-store store > + (mbegin %store-monad > +(set-guile-for-build (default-guile)) > +(case command > + ((list-generations) > + (let ((pattern (match args > +(() "") > +((pattern) pattern) > +(x (leave (_ "wrong number of arguments~%")) > + (list-generations pattern))) > + (else > + (process-action command args opts > + #:system (assoc-ref opts 'system As the comment above suggests, the idea here was to avoid connecting to the daemon for operations that do not need it, such as ‘list-generations’. I think we should preserve this property. Thanks for your work! I’m sorry this is more painful than I thought. :-/ Ludo’.
Re: Mysterious error while refactoring guix/scripts/system.scm
Hi Andy, Thank you for the help! Sorry for the late reply; I wasn't able to respond until just now. > Here the #:system system was taking the `system' binding from within the > function -- it was lexically bound. You're right; that was the problem. I've updated the patch (see attached) and applied it again to 7972d8a2e98af6592050a37036c2c80a01358fcf, but when I built the new version and ran './pre-inst-env guix system list-generations', the following new error occurred (after successfully printing the list of generations): --8<---cut here---start->8--- Generation 1Jul 29 2016 08:26:37 file name: /var/guix/profiles/system-1-link canonical file name: /gnu/store/90xaq1d04g39c1wmxplngh3hhc1h5z7g-system label: GNU with Linux-Libre 4.7 (beta) root device: /dev/sda1 kernel: /gnu/store/p7wbfpbs8p8ykdrn1440f95ap9iq1czh-linux-libre-4.7 Generation 2Aug 03 2016 07:35:00 file name: /var/guix/profiles/system-2-link canonical file name: /gnu/store/6il3znhh2cyp3xw0800x6m3yjq8jq5z9-system label: GNU with Linux-Libre 4.7 (beta) root device: my-root kernel: /gnu/store/p7wbfpbs8p8ykdrn1440f95ap9iq1czh-linux-libre-4.7 Generation 3Aug 03 2016 07:38:12 file name: /var/guix/profiles/system-3-link canonical file name: /gnu/store/74d698pk4dwv4k4bvf2n4b7j4bg0gbq0-system label: GNU with Linux-Libre 4.7 (beta) root device: #vu8(18 55 238 217 114 62 64 185 130 140 206 84 109 78 188 217) kernel: /gnu/store/p7wbfpbs8p8ykdrn1440f95ap9iq1czh-linux-libre-4.7 Generation 4Aug 03 2016 07:55:20(current) file name: /var/guix/profiles/system-4-link canonical file name: /gnu/store/01fd40gf19syn493y2f6h35cb58kqmg4-system label: GNU with Linux-Libre 4.7 (beta) root device: #vu8(18 55 238 217 114 62 64 185 130 140 206 84 109 78 188 217) kernel: /gnu/store/p7wbfpbs8p8ykdrn1440f95ap9iq1czh-linux-libre-4.7 Backtrace: In ice-9/boot-9.scm: 157: 15 [catch #t # ...] In unknown file: ?: 14 [apply-smob/1 #] In ice-9/boot-9.scm: 63: 13 [call-with-prompt prompt0 ...] In ice-9/eval.scm: 432: 12 [eval # #] In ice-9/boot-9.scm: 2401: 11 [save-module-excursion #] 4050: 10 [#] 1724: 9 [%start-stack load-stack ...] 1729: 8 [#] In unknown file: ?: 7 [primitive-load "/root/guix/scripts/guix"] In guix/ui.scm: 1209: 6 [run-guix-command system "list-generations"] In ice-9/boot-9.scm: 157: 5 [catch srfi-34 # ...] 157: 4 [catch system-error ...] In guix/scripts/system.scm: 884: 3 [#] 818: 2 [process-command list-generations () ...] In guix/store.scm: 1182: 1 [run-with-store # ...] In unknown file: ?: 0 [# #] ERROR: In procedure #: ERROR: Wrong type to apply: # --8<---cut here---end--->8--- I'm hoping you can help me (1) understand this particular problem, and (2) help me learn to become more self-sufficient at troubleshooting these kinds of issues going forward. I don't want to have to bother you or the mailing list every time I encounter a stack trace. Regarding (1), I'm having trouble understanding this particular stack trace. Where did the error occur? What caused it? I'm used to stack traces in Java and Python, where often the exact line which threw an exception is usually obvious in the stack trace. I'm finding Guile's stack traces more difficult to understand. Maybe I'm just not used to them yet. Regarding (2), I've read some of the manual at '(guile) Debugging' (mainly '(guile) Evaluation Model' and '(guile) Programmatic Error Handling'), and although it's great and detailed, it isn't obvious to me how to apply that knowledge in practice. For example, I'm not even sure how to find out what the parts and symbols in the above stack trace mean. How can I get better at this? Should I just read the rest of '(guile) Debugging' in detail, and experiment with my own toy programs until it makes more sense? Or is there a common tool people use for debugging that I'm not aware of, like how pdb is a common debugging tool for Python programs? Again, thank you for your help. I really appreciate it! -- Chris From a3e094e5393ac7289075a0b63696bd9e5024b202 Mon Sep 17 00:00:00 2001 From: Chris Marusich Date: Wed, 3 Aug 2016 00:39:39 -0700 Subject: [PATCH] Refactor process-action and process-command --- guix/scripts/system.scm | 74 - 1 file changed, 36 insertions(+), 38 deletions(-) diff --git a/guix/scripts/system.scm b/guix/scripts/system.scm index 209ebf9..731e3cd 100644 --- a/guix/scripts/system.scm +++ b/guix/scripts/system.scm @@ -774,7 +774,6 @@ resulting from command-line parsing." (let* ((file (match args (() #f) ((x . _) x))) - (system (assoc-ref opts 'system)) (os (if file (load* file %user-module #:on-error (assoc-ref opts 'on-error)) @@ -789,47 +788,46 @@ resulting from command-line parsing." (gru
Re: Mysterious error while refactoring guix/scripts/system.scm
On Thu 04 Aug 2016 08:20, Chris Marusich writes: > 1038: 3 [thunk] > In gnu/packages/bootstrap.scm: > 191: 2 [raw-build # "guile-bootstrap-2.0" ...] > In gnu/packages.scm: > 91: 1 [search-bootstrap-binary "tar" #] > In unknown file: >?: 0 [string-append # "/" "tar"] > > ERROR: In procedure string-append: > ERROR: In procedure string-append: Wrong type (expecting string): # system (#:optional _)> This is because you moved some code around: > --- a/guix/scripts/system.scm > +++ b/guix/scripts/system.scm > @@ -785,47 +785,46 @@ resulting from command-line parsing." > (grub-configuration-device > (operating-system-bootloader os) > > -(with-store store > - (set-build-options-from-command-line store opts) > - > - (run-with-store store > -(mbegin %store-monad > - (set-guile-for-build (default-guile)) > - (case action > -((extension-graph) > - (export-extension-graph os (current-output-port))) > -((shepherd-graph) > - (export-shepherd-graph os (current-output-port))) > -(else > - (perform-action action os > - #:dry-run? dry? > - #:derivations-only? (assoc-ref opts > - > 'derivations-only?) > - #:use-substitutes? (assoc-ref opts > 'substitutes?) > - #:image-size (assoc-ref opts 'image-size) > - #:full-boot? (assoc-ref opts 'full-boot?) > - #:mappings (filter-map (match-lambda > - (('file-system-mapping > . m) > - m) > - (_ #f)) > -opts) > - #:grub? grub? > - #:target target #:device device > -#:system system > +(case action > + ((extension-graph) > + (export-extension-graph os (current-output-port))) > + ((shepherd-graph) > + (export-shepherd-graph os (current-output-port))) > + (else > + (perform-action action os > + #:dry-run? dry? > + #:derivations-only? (assoc-ref opts > + 'derivations-only?) > + #:use-substitutes? (assoc-ref opts 'substitutes?) > + #:image-size (assoc-ref opts 'image-size) > + #:full-boot? (assoc-ref opts 'full-boot?) > + #:mappings (filter-map (match-lambda > +(('file-system-mapping . m) > + m) > +(_ #f)) > + opts) > + #:grub? grub? > + #:target target #:device device) Here the #:system system was taking the `system' binding from within the function -- it was lexically bound. > (define (process-command command args opts) >"Process COMMAND, one of the 'guix system' sub-commands. ARGS is its > argument list and OPTS is the option alist." > - (case command > -((list-generations) > - ;; List generations. No need to connect to the daemon, etc. > - (let ((pattern (match args > - (() "") > - ((pattern) pattern) > - (x (leave (_ "wrong number of arguments~%")) > - (list-generations pattern))) > -(else > - (process-action command args opts > + (with-store store > +(set-build-options-from-command-line store opts) > + > +(run-with-store store > + (mbegin %store-monad > +(set-guile-for-build (default-guile)) > +(case command > + ((list-generations) > + (let ((pattern (match args > +(() "") > +((pattern) pattern) > +(x (leave (_ "wrong number of arguments~%")) > + (list-generations pattern))) > + (else > + (process-action command args opts > + #:system system))) Here it's not bound in the function so we take the top-level binding, which is the `system' library call, a function. Andy
Mysterious error while refactoring guix/scripts/system.scm
Hi, In my local Guix git repo, I've made the attached change on top of commit 7972d8a2e98af6592050a37036c2c80a01358fcf. I was surprised to find that when I built it and ran './pre-inst-env guix system list-generations', the following error occurred: --8<---cut here---start->8--- Backtrace: In ice-9/boot-9.scm: 157: 19 [catch system-error ...] In guix/scripts/system.scm: 884: 18 [#] 818: 17 [process-command list-generations () ...] In guix/store.scm: 1182: 16 [run-with-store # ...] In guix/scripts/system.scm: 819: 15 [# #] In guix/packages.scm: 1092: 14 [# #] 734: 13 [cache! # # # ...] 1038: 12 [thunk] 970: 11 [bag->derivation # # #] In srfi/srfi-1.scm: 578: 10 [map # #] In guix/packages.scm: 794: 9 [expand-input # # # ...] In guix/store.scm: 1182: 8 [run-with-store # ...] In guix/packages.scm: 1140: 7 [# #] In guix/download.scm: 326: 6 [# #] In guix/store.scm: 1105: 5 [# #] In guix/packages.scm: 734: 4 [cache! # # # ...] 1038: 3 [thunk] In gnu/packages/bootstrap.scm: 191: 2 [raw-build # "guile-bootstrap-2.0" ...] In gnu/packages.scm: 91: 1 [search-bootstrap-binary "tar" #] In unknown file: ?: 0 [string-append # "/" "tar"] ERROR: In procedure string-append: ERROR: In procedure string-append: Wrong type (expecting string): # --8<---cut here---end--->8--- I don't have much experience (yet!) debugging Guile programs, so I haven't been able to understand the cause of the issue despite my attempts. Can anyone help me understand the problem, or provide tips on how to debug this more effectively? The motivation for this refactoring is as follows. I want to eventually add procedures 'roll-back-system' and 'switch-to-system-generation', which respectively roll the system back to a previous generation, and switch to an existing generation. These new procedures will require access to the store (e.g., to regenerate grub.cfg). I must invoke those procedures somewhere in (guix scripts system), and the obvious place to do that is inside the 'process-action' procedure. However, that procedure is written with the assumption that the action requires an operating system configuration file. My two new procedures don't require that. So, instead, I thought it would be easiest to pull the store monad setup logic (the 'with-store' stuff) up one level into the procedure 'process-command' and call my two new procedures inside the 'process-command' procedure. It looked to me like there was nothing specific about the store monad setup logic that required it to exist in the 'process-action' procedure, so I thought this refactoring would be simple and clean. I was apparently wrong, but I can't see why. This is the last thing blocking me from submitting patches to the email list which add system roll-back and switch-generation commands to GuixSD. I have a series of commits on a branch in my Guix git repo which implement system-rollback and switch-generation. If you'd like to see them for more context, please let me know and I'll provide them, too. Thank you! -- Chris From 1f3a08b080c75b9be7c74235637cce0f91a249d5 Mon Sep 17 00:00:00 2001 From: Chris Marusich Date: Wed, 3 Aug 2016 00:39:39 -0700 Subject: [PATCH] Refactor process-action and process-command --- guix/scripts/system.scm | 73 - 1 file changed, 36 insertions(+), 37 deletions(-) diff --git a/guix/scripts/system.scm b/guix/scripts/system.scm index e2c6b2e..8302d57 100644 --- a/guix/scripts/system.scm +++ b/guix/scripts/system.scm @@ -785,47 +785,46 @@ resulting from command-line parsing." (grub-configuration-device (operating-system-bootloader os) -(with-store store - (set-build-options-from-command-line store opts) - - (run-with-store store -(mbegin %store-monad - (set-guile-for-build (default-guile)) - (case action -((extension-graph) - (export-extension-graph os (current-output-port))) -((shepherd-graph) - (export-shepherd-graph os (current-output-port))) -(else - (perform-action action os - #:dry-run? dry? - #:derivations-only? (assoc-ref opts -'derivations-only?) - #:use-substitutes? (assoc-ref opts 'substitutes?) - #:image-size (assoc-ref opts 'image-size) - #:full-boot? (assoc-ref opts 'full-boot?) - #:mappings (filter-map (match-lambda - (('file-system-mapping . m) - m) - (_ #f)) -opts) -