Re: [libgpiod] [PATCH 5/5] bindings: cxx: Workaround --success run

[Alexander Stein <alexander.stein@mailbox.org>]

 On Thursday, August 8, 2019, 5:27:14 PM CEST Bartosz Golaszewski wrote:
> śr., 7 sie 2019 o 21:51 Alexander Stein <alexander.stein@mailbox.org>
> napisał(a):
> >
> > If run with --success, all expressions are evaluated and printed out.
> > But REQUIRE_FALSE(chip) tries to iterate over the chip resulting in this
> > backtrace:
> > #0  gpiod_chip_num_lines (chip=chip@entry=0x0) at ../../lib/core.c:235
> > #1  gpiod_line_iter_new (chip=0x0) at ../../lib/iter.c:140
> > #2  gpiod::(anonymous namespace)::make_line_iter (chip=0x0) at ../../../bindings/cxx/iter.cpp:29
> > #3  gpiod::line_iter::line_iter (this=0x7fffffffd690, owner=...) at ../../../bindings/cxx/iter.cpp:109
> > #4  Catch::rangeToString<gpiod::chip> (range=...) at /usr/include/catch2/catch.hpp:1959
> > [...]
> >
> > Workaround by forcing catch2 to call gpiod::chip::operator bool().
> >
> > Signed-off-by: Alexander Stein <alexander.stein@mailbox.org>
> > ---
> > This actually looks like a flaw in the binding itself that the
> > gpiod::line_iter can't cope with an empty gpiod::chip.
> >
> 
> Do you want to submit a patch that fixes that? Otherwise I can fix it.
> I think that simply throwing an exception on empty chip is enough,
> right?

Reading that backtrace today, the actual problem is gpiod_chip_num_lines deferencing the nullptr.
There are 2 possibilities:
* if gpiod_chip is NULL in gpiod_line_iter_new(), return NULLL iter as well (which will raise an exception on line iter.cpp:31)
* return an iter with num_lines = 0

Can't rate the 2nd one if this will raise other problems.

Best regards,
Alexander