bug#32058: mysql build fails on d88b29d6b78482cdb05ac714984f6a27195e3d37

[Ricardo Wurmus <rekado@elephly.net>]

Nils Gillmann <ng0@n0.is> writes:

> Tobias Geerinckx-Rice transcribed 5.0K bytes:
>> ng0,
>>
>> Thanks!
>>
>> Nils Gillmann wrote:
>> > Since I do not have the time to work on this, but there seems to be no
>> > open bug:
>> >
>> > mysql started a couple of commits ago (since last core-updates merge?)
>> > to fail
>> > its build.
>>
>> I noticed this yesterday, too. In the meantime, I've tried updating MySQL to
>> 5.7.22 (one never knows) and poking at some random bits but that didn't
>> help.
>>
>> > [build output snipped]
>>
>> There's actually an error message[0]. Did it not show up in your logs? I'd
>> consider that a bug too.
>
> Odd. I think this was by accident. Here it is:

This happens when building in parallel; the actual error that caused the
build to terminate could have occurred hundreds of lines earlier when
another independent make processes is still running.

Anyway, thanks for providing the error message.

> /tmp/guix-build-mysql-5.7.21.drv-0/mysql-5.7.21/sql/item_geofunc_internal.cc: In function ‘void
> handle_gis_exception(const char*)’:
> /tmp/guix-build-mysql-5.7.21.drv-0/mysql-5.7.21/sql/item_geofunc_internal.cc:37:81: error: expected
> unqualified-id before ‘&’ token
>    catch (const boost::geometry::detail::self_get_turn_points::self_ip_exception &)

The Internet tells me that mysql is very picky about which versions of
boost it build with.  We use boost 1.66.0 now.

The mysql package definition contains this comment:

   ;; Mysql wants boost-1.59.0 specifically

and then we proceed to just replace 59 for whatever minor version it is
we are using.

Should we just upgrade to the latest version 8.0.11?

--
Ricardo