Hello,

thanks to everyone involved in making the rc1 happen! :)

Olivier Dion <[email protected]> writes:

> Hello to all Guilers/Guixers,
>
> GNU Guile 3.0.12, release candidate 1 is now available.  You can either
> test it from Git using the `v3.0.12rc1' tag or from the `main' branch.
>
> Tarballs produced by `make distcheck' on that tag have also been pushed:
>
>   https://alpha.gnu.org/gnu/guile/guile-3.0.12rc1.tar.gz
>   SHA256: 1412f6f077c9f90147537778a6a902e135600db33b5d12293c5472bd635e7704
>
>   https://alpha.gnu.org/gnu/guile/guile-3.0.12rc1.tar.lz
>   SHA256: b097447b8c5ee5505522c8dbe21e88db6079a90551ae17c5b3005ae53d147a1b
>
>   https://alpha.gnu.org/gnu/guile/guile-3.0.12rc1.tar.xz
>   SHA256: 1bcd59d333af9b35c5ede07f2cff174d46b205518f48d4327d870a0aaf0b2828
>
> Each tarball has a corresponding detached signature file (`.sig') made
> with the following PGP key: 295C02464AC192F1FFDD7550FCC088CE07A04DAE
>
> Note that the The SONAME of `libguile-3.0.so' is unchanged, and the full
> library version was bumped from `1.7.1' to `1.8.0'.
>
> Also note that the object code version was bumped from `4.7' to `4.8'
> (this is not the case on `v3.0.12rc1' but it is on `main').  This means
> that bytecodes (`.go' files) from `3.0.11' are still backward compatible
> with `3.0.12'.  However, bytecodes of `3.0.12' can not be loaded by
> `3.0.11'.  So don't be suprise if you get a bunch of `auto-compile'
> messages if you are loading bytecodes from `3.0.12' with older Guile.
>
> The release candidate cycle start today (Nov. 17th) and will end,
> hopefully, mid-december.  If all goes well, I was hoping of making the
> official release before the holidays.  Until then, what I mostly need
> from you all is to test as much as possible and to report back any
> regression you have.  I will not be fixing new issues nor adding
> enhancement or doc fixes until the release.  I will only be fixing
> regressions that were introduced during this release.

Test suite of my personal library fully passes on the 3.0.12rc1, FWIW.

>
> Finally, note that I did some improvement on the message reporting side
> of the compiler.  Mainly in the reader and on the CPS analyze passes.
> If your program/tests rely on message layout for these, this could be
> considered a regression.  However, I think that these improvement are
> very important and much welcomed by the users.  So if you can add
> workarounds for this kind of regression, that would be great.

I agree the new warnings/errors look pretty and are probably useful for
interactive use in a terminal, however in CI they complicate suppressing
warnings (due to every message being multiple lines).

My current thinking is basically to put following grep before the one I
already have, but that is prone to false positives.

    grep -vEe '^ +| +\^~+' -e '^ +[0-9]+ +|'

Would it be possible to get either environment variable or command line
flag to keep the one-line format?  I am not asking for it being default,
but it would be nice to have a more robust way of doing this.

Or the subsequent lines could have harmless, likely-to-be-unique prefix,
like \b.

Tomas

-- 
There are only two hard things in Computer Science:
cache invalidation, naming things and off-by-one errors.

Reply via email to