bug#81490: 32.0.50; wrong-type-argument internal to package-install

Mats Lidell <[email protected]> writes:

> Hi,
>
> In Hyperbole we run a CI workflow in Github for Emacs 28.2, 29.4, 30.2
> and master based on the docker images produced by
> https://github.com/Silex/docker-emacs. Hyperbole has a dynamic
> dependency on markdown-mode and installs it when needed in the CI
> workflow. That happens during test execution, ert tests.
>
> A week ago we changed the code and then we got into this problem. For
> the master build we get the backtrace below. The error seems to be
> related to a race condition. Sometimes it works but mostly it
> fails. Only master seems to have this problem. We have not seen this
> for 28.2, 29.4 and 30.2. Unfortunately I can't recreate the issue
> running docker locally using the same master image as in Github.

Do you know if this issue also occurs with the pre-release Emacs 31?

> As a workaround in the CI workflow we have for Emacs master (32.0.50)
> adviced package-install to retry with a one second delay if it signals
> wrong-type-argument. Then it works on the second attempt.
>
> Backtrace:
> ----------------------------------------------------------------------
>  signal((wrong-type-argument stringp nil))
>   package--with-response-buffer-1("https://elpa.nongnu.org/nongnu/" #f(compiled-function () #<bytecode 0x1e93d79faeaf8db2>) :file "markdown-mode-2.8.tar" :async nil :error-function #f(compiled-function () #<bytecode 0x18c000af2941>) :noerror nil)
>   package-install-from-archive(#s(package-desc :name markdown-mode :version (2 8) :summary "Major mode for Markdown-formatted text" :reqs (...) :kind tar :archive "nongnu" :dir nil :extras (... ... ... ... ...) :signed nil))
>   package-download-transaction((#s(package-desc :name markdown-mode :version (2 8) :summary "Major mode for Markdown-formatted text" :reqs (...) :kind tar :archive "nongnu" :dir nil :extras (... ... ... ... ...) :signed nil)))
>   package-install(markdown-mode)
>   hypb:package-el-install(markdown-mode)
>   [...]
> ----------------------------------------------------------------------

This backtrace gives me no indication where exactly this bug comes from,
and looking at the parameter list we are not passing any suspicious
values to the function.  Are you sure that this is not an issue with the
Docker container?

> Yours