Re: mail/notmuch build fails on 9.0_RC1

[Greg Troxel <gdt%lexort.com@localhost>]

Carolyn Knight-Serrano <gigavinyl%riseup.net@localhost> writes:

> Howdy! I'm trying to build mail/notmuch on 9.0_RC1 but I'm getting a
> strange xapian error:
>
> ld: lib/libnotmuch.a(thread-fp.o): in function
> `ThreadFieldProcessor::operator()(std::__cxx11::basic_string<char,
> std::char_traits<char>, std::allocator<char> > const&)':
> thread-fp.cc:(.text+0x296): undefined reference to
> `Xapian::Query::Query(std::__cxx11::basic_string<char,
> std::char_traits<char>, std::allocator<char> > const&, unsigned int,
> unsigned int)'
> ld: thread-fp.cc:(.text+0x34b): undefined reference to
> `Xapian::Query::Query(std::__cxx11::basic_string<char,
> std::char_traits<char>, std::allocator<char> > const&, unsigned int,
> unsigned int)'
> gmake: *** [Makefile.local:257: notmuch] Error 1
> *** Error code 2
>
> Should this package be updated/marked as broken? Thanks!

Are you using pkgsrc current, or 2019Q3?

Are you really sure that you have built everything with the same
compiler version?   C++ build issues are often tricky to track down, but
seeing std::__cxx11 show up in undefined symbols makes me wonder if the
xapian library was built with an earlier compiler (under 8?) and notmuch
is being built under 9.

I just built notmuch on netbsd-8, and other than not having RPATH
encoded in the binary, it worked.

That doesn't mean there isn't bug on -9, but if so it should be fixable.

I'd suggest

  pkg_info -B xapian|egrep CC_V
  cc -v

and if not matching, make replace in /usr/pkgsrc/textproc/xapian.
On my system, both are 5.5.0.

To be more thorough, you might also install pkg_rolling-replace

  pkg_admin set rebuild=YES \*
  pkg_rolling-replace -uvk

or, make a list of packages (e.g. with pkg_chk), remove them all, and
build them all up.

Otherwise, someone will need to understand the C++ details, or more
likely sooner, see what happens on their -9 system.

A bulk build report from a few days ago shows notmuch building, or more
specifically doesn't show it in the failed list.