Re: [boost] [test] --catch_system_errors=no doesn't work withminimaltest

From: David Abrahams (dave_at_[hidden])
Date: 2005-11-01 16:11:13

"Gennadiy Rozental" <gennadiy.rozental_at_[hidden]> writes:

> "David Abrahams" <dave_at_[hidden]> wrote in message 
> news:uu0ew8rxm.fsf_at_boost-consulting.com...
>> "Gennadiy Rozental" <gennadiy.rozental_at_[hidden]> writes:
>>
>>>> When using boost/test/minimal.hpp, it doesn't respond to the presence
>>>> of --catch_system_errors=no, which makes it unreasonably difficult to
>>>> debug problems on Windows.  All I get is an exception caught by the
>>>> test library, with all of the info about the real cause of the error
>>>> lost, which is absolutely unhelpful.  It seems to me that the minimal
>>>> configuration ought to at least be useful for writing and debugging
>>>> simple test cases.
>>>
>>> Minimal testing facility doesn't process any CLA.
>>
>> http://boost.org/libs/test/doc/usage/command_line.html and
>> http://boost.org/libs/test/doc/usage/msvc_net.html#t2 don't make that
>> clear.
>
> Minimal testing facility is in many sences "crippled" relative to te rest of 
> Boost.Test components. There is a separate chapter covering it and the rest 
> of docs doesn't have anything to do with it. If you think it worth the 
> effort I could explicetly specify in all places: "Is not applicable ot 
> Minimal Testing facility"

Either that, or make it very very clear at the beginning of the
minimal test docs that no other library documentation is valid for the
minimal facility.

>>> Execution monitor by default catches all the errors (which I think
>>> reasonable default).
>>
>> I'm not sure it is.  It seems to me that Boost.Test has two uses:
>>
>>  1) for writing effective unit and regression tests
>>  2) for writing "resilient" software
>>
>> Given its name, I think (1) should be emphasized, but I think the
>> "catches all errors" behavior is more appropriate to (2).
>
> I guess it's a matter of personal preferense either. I nmany cases I
> don't need core to fix an error anonced with "fatal error"
> message. And if do need I could always specify extra parameter. 

What "extra parameter?"  Where?
If you mean "command line **argument**," then it doesn't apply because
minimal.hpp, for some reason, ignores them.

> On the other hand regression tests runners never need this. 

They do if they are asked to debug a problem that nobody else can
reproduce.  It happens occasionally.

> And I wouldn't want to force them to add this parameter always. IOW
> I think we need more compeling reason to change a default.

Well, when people are just dipping their toes in the test library
waters they are likely to start with minimal.hpp, and they are likely
to be trying to do some local testing rather than writing some test
that will be run by a 3rd party under some large test harness.  It
seems to me that all your reasons for not changing it are directed at
that case, though.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com