DMD unittest fail reporting…

[Chris Wright via Digitalmars-d]

On Fri, 04 Dec 2015 19:00:37 +0000, Russel Winder via Digitalmars-d wrote:

> … is completely hideous, or am I unique in objecting to the mess of
> output you get on a test fail?

You are probably thinking mainly of failing assertions directly in the 
body of a unittest block, but assertions can happen anywhere, and 
uncaught exceptions can also fail a unittest. In either of these cases, I 
do want to see a stacktrace.

The problem with stacktraces at the moment is this:

core.exception.AssertError at source/url.d(925): Assertion failure
----------------
??:? _d_assert [0x52b10f]
??:? void url.__assert(int) [0x528597]
source/url.d:925 immutable(char)[] url.punyDecode(immutable(char)[]) 
[0x5139b0]
source/url.d:989 void url.__unittestL988_12() [0x513c39]
??:? void url.__modtest() [0x528535]
??:? int core.runtime.runModuleUnitTests().__foreachbody2
(object.ModuleInfo*) [0x561c3a]
??:? int object.ModuleInfo.opApply(scope int delegate
(object.ModuleInfo*)).__lambda2(immutable(object.ModuleInfo*)) [0x52a8d7]
??:? int rt.minfo.moduleinfos_apply(scope int delegate(immutable
(object.ModuleInfo*))).__foreachbody2(ref rt.sections_elf_shared.DSO) 
[0x53137a]
??:? int rt.sections_elf_shared.DSO.opApply(scope int delegate(ref 
rt.sections_elf_shared.DSO)) [0x531409]
??:? int rt.minfo.moduleinfos_apply(scope int delegate(immutable
(object.ModuleInfo*))) [0x53130b]
??:? int object.ModuleInfo.opApply(scope int delegate(object.ModuleInfo*)) 
[0x52a8b3]
??:? runModuleUnitTests [0x561aad]
??:? void rt.dmain2._d_run_main(int, char**, extern (C) int function(char
[][])*).runAll() [0x52da06]
??:? void rt.dmain2._d_run_main(int, char**, extern (C) int function(char
[][])*).tryExec(scope void delegate()) [0x52d9b4]
??:? _d_run_main [0x52d911]
??:? main [0x50ee8f]
??:? __libc_start_main [0xcb7e2ec4]

20 lines of stacktrace. The potentially useful portion of the stacktrace:

source/url.d:925 immutable(char)[] url.punyDecode(immutable(char)[]) 
[0x5139b0]
source/url.d:989 void url.__unittestL988_12() [0x513c39]

I think a good approximation would be to skip from the head of the stack 
to the first non-runtime frame, but, if no non-runtime frames exist, emit 
the whole stacktrace. That wouldn't hide anything in the case of a 
runtime crash, but it would eliminate about 15 lines of cruft.