On Tue, 15 Sep 2020 14:26:48 +0100
Ben Hutchings <ben@xxxxxxxxxxxxxxx> wrote:
> On Tue, 2020-09-15 at 16:12 +0300, Tzvetomir Stoyanov wrote:
> > On Tue, Sep 15, 2020 at 12:29 AM Arnaldo Carvalho de Melo
> > <arnaldo.melo@xxxxxxxxx> wrote:
> > > Em Fri, Sep 11, 2020 at 03:08:16AM +0100, Ben Hutchings escreveu:
> > > > I noticed that the new function tep_free_plugin_paths() is exported
> > > > from libtraceevent, but is only declared in a private header file.
> >
> > Hi Ben,
> > The tep_free_plugin_paths() function is supposed to be an internal
> > function, not an API - that's why it is in a private header. What do you
> > mean by "exported": the "tep_" prefix, or the fact that it is not static?
>
> It is exported from the shared library, i.e. it's included in the
> library's dynamic symbols. That means that it could be called by a
> program using the library, and would conflict with a function defined
> by the program.
Of course a program that uses this shouldn't have any tep_ prefix
functions ;-)
>
> > I can remove the prefix, if that is what bothers you.
>
> No, exporting symbols without that prefix would make conflicts more
> likely.
Right!
>
> > The function is
> > defined in event-plugin.c and is used in event-parse.c, that's why it
> > cannot be made static without a library redesign.
>
> Yes, I realise that. However you can define it as "hidden":
> <https://gcc.gnu.org/onlinedocs/gcc-10.2.0/gcc/Common-Function-Attributes.html#index-visibility-function-attribute>.
Ah, that's how libraries can hide non static functions. I learned something
new today!
Tzvetomir,
Let's only have functions that are part of the library's API start with
"tep_", and all other functions not have "tep_" in the name (so we know
which ones are exported and which ones are not).
Then any function that doesn't start with "tep_" should either be static,
or hidden.
#define __hidden __attribute__((visibility ("hidden")))
static void foo ()
{
}
void __hidden bar ()
{
}
-- Steve
|