Re: [RFC v3 03/19] kunit: test: add string_stream a std::stream like string builder

From: Brendan Higgins
Date: Mon Dec 03 2018 - 19:35:27 EST


On Mon, Dec 3, 2018 at 2:55 AM Petr Mladek <pmladek@xxxxxxxx> wrote:
>
> On Thu 2018-11-29 19:29:24, Luis Chamberlain wrote:
> > On Wed, Nov 28, 2018 at 11:36:20AM -0800, Brendan Higgins wrote:
> > > A number of test features need to do pretty complicated string printing
> > > where it may not be possible to rely on a single preallocated string
> > > with parameters.
> > >
> > > So provide a library for constructing the string as you go similar to
> > > C++'s std::string.
> >
> > Hrm, what's the potential for such thing actually being eventually
> > generically useful for printk folks, I wonder? Petr?
>
> printk() is a bit tricky:
>
> + It should work in any context. Any additional lock adds risk of a
> deadlock. Especially the NMI and scheduler contexts are problematic.
> There are problems with any other code that might be called
> from console drivers and calls printk() under a lock.
>
> + It should work also when the system is out of memory. Especially
> atomic context is problematic because we could not wait for
> memory reclaim or swap.
>
> + We also do to the best effort to get the message out on the
> console. It is important when the system is about to die.
> Any extra buffering layer might cause delay and avoid seeing the
> message.
>
> From this point of views, this API is not generally usable with printk().

Yeah, that makes sense. I wouldn't really expect this to work well in
those cases.

> Now, the question is how many of the above fits also for unit testing.
> At least, you might need to be careful when allocating memory in
> atomic context.

True, but this is only supposed to be used for constructing
expectation failure messages which should only happen from a
non-atomic context.

>
> BTW: There are more existing printk APIs: Well, I admit the they are
> not easily reusable in unit testing:
>
> + printk() is old, crappy code, complicated with all the
> cornercases and consoles.
>
> + include/linux/seq_buf.h is simple buffering. It is used primary
> for sysfs output. It might be usable if you add support for
> loglevel and use big enough buffer. I quess that you should
> flush the buffer regularly anyway.
>
> + trace_printk() uses lockless per-CPU buffers. It currently does not
> support loglevels. But it might be pretty interesting choice as well.
>
>
> I do not say that you have to use one of the existing API. But you
> might consider them if you encouter any problems and maintaining
> your variant gets complicated.

Alright, I will take a look.

Thanks!