On Wed, May 5, 2021 at 7:09 AM Havard Skinnemoen <[email protected]> wrote:
> On Wed, May 5, 2021 at 6:55 AM Peter Maydell <[email protected]> > wrote: > > > > In the glib API, the distinction between g_assert() and > > g_assert_true() is that the former is for "bug, terminate the > > application" and the latter is for "test check, on failure either > > terminate or just mark the testcase as failed". For QEMU, g_assert() > > is always fatal, so code can assume that if the assertion fails > > execution does not proceed, but this is not true of g_assert_true(). > > > > In npcm7xx_pwm-test, the pwm_index() and pwm_module_index() functions > > include some assertions that are just guarding against possible bugs > > in the test code that might lead us to out-of-bounds array accesses. > > These should use g_assert() because they aren't part of what the test > > is testing and the code does not correctly handle the case where the > > condition was false. > > > > This fixes some Coverity issues where Coverity knows that > > g_assert_true() can continue when the condition is false and > > complains about the possible array overrun at various callsites. > > > > Fixes: Coverity CID 1442340, 1442341, 1442343, 1442344, 1442345, 1442346 > > Signed-off-by: Peter Maydell <[email protected]> > > Reviewed-by: Havard Skinnemoen <[email protected]> > Reviewed-by: Hao Wu <[email protected]>
