"Unable to execute port" when trying to install pandoc from source

[szhorvat (Szabolcs Horvát)]

When trying to install pandoc from source, I get:

$ sudo port -s install pandoc
--->  Computing dependencies for pandocError: Unable to execute port: can't read "configure.env": can't read "haskell_stack.env": can't read "haskell_stack.yaml": can't read "worksrcpath": can't read "worksrcdir": can't read "distname": can't read "name": no such variable

I attempted to update it to 2.14.0.1, but because of this problem, I was not able to.

[essandess (Steve Smith)]

I am not able to replicate this issue.

Did you start with a basic wipe of previous builds?

sudo port clean --all pandoc

[szhorvat (Szabolcs Horvát)]

Yes, I did. I will attach the output of

sudo port -vd install pandoc

but note that this one was run with the portfile changed in an attempt to update to 2.14.0.1 (however, only the version/checksum was changed).

[essandess (Steve Smith)]

I do not observe this issue. Please see ​https://github.com/macports/macports-ports/pull/11315

[szhorvat (Szabolcs Horvát)]

Thanks for submitting the update!

It seems that the CI checks do fail for the PR. While I found it difficult to make sense of the CI output, if I download the "artifacts" here:

​https://github.com/macports/macports-ports/pull/11315/checks?check_run_id=2803884619

and look at install-dependencies.log, I see the same error messages that I get on my own machine.

Do you have a fully updated MacPorts on your computer?

[jmroot (Joshua Root)]

Seems to be an interaction between the legacysupport and haskell_stack portgroups. The former reads configure.env at the top level, which is evaluated before name is set, and the latter sets the default value of configure.env in a way that depends on name.

[Chris Jones <jonesc@…>]

In 4d0ef085ac02125699c90ce0388105ecc2dec978/macports-ports (master):

legacysupport PG: Revert configuring when PG is first included
Closes: #63076

[cjones051073 (Chris Jones)]

I recently update the PG and part of that was to configure as soon as the PG was included, as well as via a callback. It seems this was not a good idea so I just reverted this one change.