Re: Buildbot failure in Wildebeest Builder on whole buildset

[Mark Wielaard <mark@klomp.org>]

Hi Philip,

On Fri, Aug 06, 2021 at 03:31:07PM +0100, Philip Herron wrote:
> > Is it possible to make bors check all commits in a series? Or can we
> > somehow connect the buildbot workers to the bors checks?
> 
> I think the build-bot checks for every commit is a really nice feature
> here. I think you should keep the build-bots as are. I really like to
> make sure each of my commits are build-able and pass tests. It keeps me
> honest to avoid regressions as best I can.
> 
> What do you think?

I am just a little afraid it is running too late. But I am pretty
happy we have workers for 4 different architectures and they are green
most of the time. We really should add a 32bit (i686 or armhf) one
just to be sure that works correctly.

Currently i686 gets lots of failures like:

/home/mark/gccrs/gcc/testsuite/rust/compile/torture/methods3.rs:27:5: error: type mismatch in binary exp
ression
f64
<float:80>
<float:80>
D.229 = _2 + _4;
/home/mark/gccrs/gcc/testsuite/rust/compile/torture/methods3.rs:27:5: internal compiler error: 'verify_g
imple' failed
0x8b41fd4 verify_gimple_in_seq(gimple*)
        ../../gccrs/gcc/tree-cfg.c:5157
0x884b4c3 gimplify_body(tree_node*, bool)
        ../../gccrs/gcc/gimplify.c:15401
0x884b692 gimplify_function_tree(tree_node*)
        ../../gccrs/gcc/gimplify.c:15472
0x86a6448 cgraph_node::analyze()
        ../../gccrs/gcc/cgraphunit.c:670
0x86a9218 analyze_functions
        ../../gccrs/gcc/cgraphunit.c:1236
0x86a9e01 symbol_table::finalize_compilation_unit()
        ../../gccrs/gcc/cgraphunit.c:2514

Cheers,

Mark