Re: [PATCH 3/3] Use g_new() & friends where that makes obvious sense

[Markus Armbruster <armbru@xxxxxxxxxx>]

Philippe Mathieu-Daudé <philippe.mathieu.daude@xxxxxxxxx> writes:

> On 15/3/22 14:59, Markus Armbruster wrote:
>> Alex Bennée <alex.bennee@xxxxxxxxxx> writes:
>> 
>>> Markus Armbruster <armbru@xxxxxxxxxx> writes:
>>>
>>>> g_new(T, n) is neater than g_malloc(sizeof(T) * n).  It's also safer,
>>>> for two reasons.  One, it catches multiplication overflowing size_t.
>>>> Two, it returns T * rather than void *, which lets the compiler catch
>>>> more type errors.
>>>>
>>> <snip>
>>>> diff --git a/semihosting/config.c b/semihosting/config.c
>>>> index 137171b717..6d48ec9566 100644
>>>> --- a/semihosting/config.c
>>>> +++ b/semihosting/config.c
>>>> @@ -98,7 +98,7 @@ static int add_semihosting_arg(void *opaque,
>>>>       if (strcmp(name, "arg") == 0) {
>>>>           s->argc++;
>>>>           /* one extra element as g_strjoinv() expects NULL-terminated 
>>>> array */
>>>> -        s->argv = g_realloc(s->argv, (s->argc + 1) * sizeof(void *));
>>>> +        s->argv = g_renew(void *, s->argv, s->argc + 1);
>>>
>>> This did indeed break CI because s->argv is an array of *char:
>>>
>>> ../semihosting/config.c:101:17: error: assignment to ‘const char **’ from 
>>> incompatible pointer type ‘void **’ [-Werror=incompatible-pointer-types]
>>>    101 |         s->argv = g_renew(void *, s->argv, s->argc + 1);
>>>        |                 ^
>>> cc1: all warnings being treated as errors
>>>
>>> So it did the job of type checking but failed to build ;-)
>>
>> You found a hole in my compile testing, thanks!
>>
>> I got confused about the configuration of my build trees.  Catching such
>> mistakes is what CI is for :)
>
> FYI Alex fixed this here:
> https://lore.kernel.org/qemu-devel/20220315121251.2280317-8-alex.bennee@xxxxxxxxxx/
>
> So your series could go on top (modulo the Coverity change).

I dropped this hunk in v2.

Whether my v2 or Alex's series goes in first doesn't matter.