On Mon, Feb 25, 2019 at 12:07:22PM -0300, Arnaldo Carvalho de Melo wrote:
SNIP
> > > > +}
> > > > +
> > > > static int open_file_read(struct perf_data *data)
> > > > {
> > > > struct stat st;
> > > > @@ -254,6 +270,22 @@ static int open_file_dup(struct perf_data *data)
> > > > return open_file(data);
> > > > }
> > > >
> > > > +static int open_dir(struct perf_data *data)
> > > > +{
> > > > + if (perf_data__is_write(data) &&
> > > > + mkdir(data->path, S_IRWXU) < 0)
> > > > + return -1;
> > > > +
> > > > + /*
> > > > + * So far we open only the header, so we
> > > > + * can read the data version and layout.
> > > > + */
> > > > + if (asprintf(&data->file.path, "%s/header", data->path) < 0)
> > > > + return -ENOMEM;
> > >
> > > so, if this fails, then we should unwind the mkdir, if it was
> > > performed, so that we leave things as they were before calling
> > > open_dir(), right?
> >
> > I think we need to some global solution on this,
>
> If we don't add more, that would be a good thing :-)
>
> The other parts also need to be investigated to see what is best in that
> case, but here, undoing the mkdir() if the asprintf() fails is the right
> thing to do :-)
true ;-) will resend
thanks,
jirka
