| Message ID | 20210924110506.4676-1-attila@lendvai.name |
|---|---|
| State | Accepted |
| Headers | show |
| Series | [bug#50751,1/2] gnu: python-trezor-agent: Fix key generation with trezor-gpg init | expand |
| Context | Check | Description |
|---|---|---|
| cbaines/comparison | success | View comparision |
| cbaines/git branch | success | View Git branch |
| cbaines/applying patch | success | View Laminar job |
| cbaines/issue | success | View issue |
| cbaines/comparison | success | View comparision |
| cbaines/git branch | success | View Git branch |
| cbaines/applying patch | success | View Laminar job |
| cbaines/issue | success | View issue |
Hello, Attila Lendvai <attila@lendvai.name> writes: > Someone with a better understanding of the python infrastructure may be able > to fix it nicer, but it works. > > * gnu/packages/finance.scm (trezor-agent): Add comment on why the > undoing of the wrapping is done, and also delete the irrelevant sideffect of > the now undone wrapping. > (python-trezor-agent): Add a patch that changes the python code to handle the > argv[0] changed by the wrapping. > * gnu/packages/patches/trezor-agent-fix-argv0.patch: New file. I prefixed the patch name with python-... and referenced it in local.mk. Then I applied it. Thank you. Regards,
ouch, i already had a newer version of the patchset locally. i reported an issue upstream, and they got rid of the argv[0] usage, and also made a new release. i was already testing it locally but then life happened before i could send it. i have sent the new patchset now: https://issues.guix.gnu.org/52207 re splitting the commit: i don't think it's a good idea, mostly due to git bisect: pushing commits that may not even compile can make someone's life much harder down the road. i think updating tightly interdependent packages should happen in one commit, so that each git commit leaves the repo in a functional state, and so that it can compile and run the test suite. but in spite of the above, i have sent the new patchset as separate commits. - attila PGP: 5D5F 45C7 DFCD 0A39
Hello, Attila Lendvai <attila@lendvai.name> writes: > ouch, i already had a newer version of the patchset locally. > i reported an issue upstream, and they got rid of the argv[0] usage, > and also made a new release. It happens. > > i was already testing it locally but then life happened before i could > send it. i have sent the new patchset now: > https://issues.guix.gnu.org/52207 Great. > re splitting the commit: i don't think it's a good idea, mostly due to > git bisect: pushing commits that may not even compile can make > someone's life much harder down the road. i think updating tightly > interdependent packages should happen in one commit, so that each git > commit leaves the repo in a functional state, and so that it can > compile and run the test suite. > > but in spite of the above, i have sent the new patchset as separate > commits. I overlooked the fact that packages were interdependent, which is pretty obvious now you mention it. In that case, I agree splitting the commit was not a good idea. So... could you send the new patchset again in a single commit? Thanks, Regards,
diff --git a/gnu/packages/finance.scm b/gnu/packages/finance.scm index 9b073541de..a123efe207 100644 --- a/gnu/packages/finance.scm +++ b/gnu/packages/finance.scm @@ -844,7 +844,8 @@ the Monero GUI client.") (commit (string-append "v" version)))) (file-name (git-file-name name version)) (sha256 - (base32 "0q99vbfd3h85s8rnjipnmldixabqmmlk5w9karv6f0rhyi54f4zv")))) + (base32 "0q99vbfd3h85s8rnjipnmldixabqmmlk5w9karv6f0rhyi54f4zv")) + (patches (search-patches "trezor-agent-fix-argv0.patch")))) (build-system python-build-system) (arguments `(#:phases @@ -1089,7 +1090,15 @@ the KeepKey Hardware Wallet.") (add-after 'wrap 'fixup-agent-py (lambda* (#:key inputs outputs #:allow-other-keys) (let* ((out (assoc-ref outputs "out"))) - ;; overwrite the wrapper with the real thing. + ;; The wrap phase also wraps trezor_agent.py (besides the + ;; public facing executable called trezor-agent). We need to + ;; undo that wrapping. The reason this is needed is that the + ;; python easy install generates a toplevel script (?) that + ;; messes with argv[0] and then re-opens the python + ;; module. This fails when the wrapped file is actually a shell + ;; script, not a python file. + (delete-file (string-append out "/bin/.trezor_agent.py-real")) + ;; Overwrite the wrapped one with the real thing. (install-file "./trezor_agent.py" (string-append out "/bin")) #t)))))) diff --git a/gnu/packages/patches/trezor-agent-fix-argv0.patch b/gnu/packages/patches/trezor-agent-fix-argv0.patch new file mode 100644 index 0000000000..9462067cd5 --- /dev/null +++ b/gnu/packages/patches/trezor-agent-fix-argv0.patch @@ -0,0 +1,27 @@ +diff --git a/libagent/gpg/__init__.py b/libagent/gpg/__init__.py +index 3711bc8..67085de 100644 +--- a/libagent/gpg/__init__.py ++++ b/libagent/gpg/__init__.py +@@ -122,15 +122,19 @@ def run_init(device_type, args): + verify_gpg_version() + + # Prepare new GPG home directory for hardware-based identity +- device_name = os.path.basename(sys.argv[0]).rsplit('-', 1)[0] +- log.info('device name: %s', device_name) ++ exe_name = os.path.basename(sys.argv[0]) ++ # drop the Guix wrapper's dot prefix from the name ++ if exe_name[0] == '.' and exe_name.endswith('-real'): ++ exe_name = exe_name[1:-5:] ++ device_name = exe_name.rsplit('-', 1)[0] ++ log.info('exe name: %s, device name: %s', exe_name, device_name) + homedir = args.homedir + if not homedir: + homedir = os.path.expanduser('~/.gnupg/{}'.format(device_name)) + + log.info('GPG home directory: %s', homedir) + +- if os.path.exists(homedir): ++ if os.path.exists(homedir) and not args.subkey: + log.error('GPG home directory %s exists, ' + 'remove it manually if required', homedir) + sys.exit(1)