Re: lang/rust PLIST fix

["Johnny C. Lam" <jlam%NetBSD.org@localhost>]

On Sun, Oct 01, 2017 at 10:59:30AM -0400, Gary Duzan wrote:
> In Message <20171001142017.7D4A111535B%xen1.duzan.org@localhost>,
>    "Gary Duzan" <gary%duzan.org@localhost>wrote:
> 
> =>   I just managed to get lang/rust to build on amd64 with the latest
> =>netbsd-8 branch, but I had to do the following:
> =>
> =>	* Apply a pthread_attr fix from -current (which has been
> =>	  requested for pullup.)
> =>	* Apply the included patch to the PLIST.
> =>
> =>   The first issue should be addressed once the pullup goes through,
> =>but I'm concerned about the second going forward. While I did make
> =>one change to fix a misapplied global replace, the bulk of the
> =>changes relate to the hashes in the filenames, and presumably that
> =>could change fairly often, making maintenance a pain. It looks like
> =>this has come up on OpenBSD as well, and there may be recent rust
> =>changes to make the hashes more stable:
> =>
> =>	https://github.com/rust-lang/rust/issues/39786
> =>	https://github.com/rust-lang/cargo/pull/3611
> =>
> =>   However, even with that change new rust release would still
> =>require a significant PLIST update. Has anyone looked into doing
> =>a dynamic PLIST for rust? I could take a look, but I don't have a
> =>particular background with rust; I was just trying to get the latest
> =>firefox built.
> 
>    Of course, right after sending this I saw from the history
> that there used to be ${RUST_VERHASH} all over the PLIST, but a
> recent commit got rid of them.

Is there a known algorithm that the rust build process uses to
generate the hash values?  Can we just duplicate it in a
GENERATE_PLIST script fragment in the rust package Makefile?

Regards,
-- 
Johnny C. Lam
jlam%NetBSD.org@localhost