On Wed, Mar 24, 2021 at 11:53 PM Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>
> On Wed, 24 Mar 2021 15:51:43 +0200
> "Tzvetomir Stoyanov (VMware)" <tz.stoyanov@xxxxxxxxx> wrote:
>
> > @@ -366,36 +366,32 @@ void __noreturn __vdie(const char *fmt, va_list ap)
> > vfprintf(stderr, fmt, ap);
> >
> > fprintf(stderr, "\n");
> > - exit(ret);
> > + return ret;
> > }
> >
> > -void __noreturn __die(const char *fmt, ...)
> > +#ifdef LIB_DIE
>
> I wonder if it would be more useful not to have it as a compile time
> define, but instead a global variable:
>
> static bool lib_die;
>
> void tracecmd_die_on_error(void)
> {
> lib_die = true;
> }
>
There is tracecmd_set_debug() API already, It can be used to trigger
that die() logic. It makes sense for the library to do a real die()
when running in debug mode, if a fatal error is encountered.
> Make the "die()" functions into static variables here.
>
> void __hidden _lib_fatal(const char *fmt, ...)
> {
> va_list ap;
>
> va_start(ap, fmt);
> if (lib_die)
> __vdie(fmt, ap);
> else
> __vwarning(fmt, ap);
> va_end(ap);
> }
>
> That way if we run trace-cmd with the --debug option, it could then call
> the tracecmd_die_on_error() function, and it will crash on any "fatal"
> places.
>
> -- Steve
>
>
> > +
> > +void __noreturn _lib_fatal(const char *fmt, ...)
> > {
> > va_list ap;
> > + int ret;
> >
> > va_start(ap, fmt);
> > - __vdie(fmt, ap);
> > + ret = __vlib_fatal(fmt, ap);
> > va_end(ap);
> > + exit(ret);
> > }
> >
> > -void __weak __noreturn die(const char *fmt, ...)
> > +#else
> > +void _lib_fatal(const char *fmt, ...)
> > {
> > va_list ap;
> >
> > va_start(ap, fmt);
> > - __vdie(fmt, ap);
> > + __vlib_fatal(fmt, ap);
> > va_end(ap);
> > }
> > -
> > -void __weak *malloc_or_die(unsigned int size)
> > -{
> > - void *data;
> > -
> > - data = malloc(size);
> > - if (!data)
> > - die("malloc");
> > - return data;
> > -}
> > +#endif
> >
--
Tzvetomir (Ceco) Stoyanov
VMware Open Source Technology Center
|