On Sat, 28 Dec 2024 23:38:49 +0200
Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote:
> On Sat, Dec 28, 2024 at 06:43:28PM +0000, David Laight wrote:
> > Commit ceb8bf2ceaa77 ("module: Convert default symbol namespace to string
> > literal") changed DEFAULT_SYMBOL_NAMESPACE to be a string literal.
> > However the conditional definition of _EXPORT_SYMBOL() was left in.
> >
> > Instead just default DEFAULT_SYMBOL_NAMESPACE to "" and remove the
> > extra _EXPORT_SYMBOL() wrapper.
> >
> > This lets DEFAULT_SYMBOL_NAMESPACE be defined after export.h is included.
>
> > Fixes fd57a3325a779 ("i2c: designware: Move exports to I2C_DW namespaces")
>
> Incorrect format, and this should be a tag.
Except that it doesn't want to be picked up by the back-port bots.
At least not in this form - since the changes to export.h that remove
the __stringify() don't really want back-porting.
>
> ...
>
> This patch in a different form had been already submitted by Uwe.
Did that move the DEFAULT_SYMBOL_NAMESPACE define to the top of the file?
That can be back-ported provided the " are removed.
The simplification to export.h (which is what I was trying to do)
can then be done after the other patches.
> So, guys, fix
> the documentation or clarify it and when you agree on the approach, choose the
> patch to review. No Ack till that. Andi, FYI.
>
I had another thought overnight - which is more changes.
Instead of using DEFAULT_SYMBOL_NAMESPACE in EXPORT_SYMBOL() use it as
a default for EXPORT_SYMBOL_NS().
So you have something like:
#define EXPORT_SYMBOL_NS(sym, ...) _EXPORT_SYMBOL_NS(sym, __VA_ARGS__, DEFAULT_SYMBOL_NAMESPACE)
#define _EXPORT_SYMBOL_NS(sym, ns, ...) __EXPORT_SYMBOL(sym, "", ns)
That requires that all the EXPORT_SYMBOL(sym) in files that define DEFAULT_SYMBOL_NAMESPACE
be changed to EXPORT_SYMBOL_NS(sym).
But it doesn't require a default definition of DEFAULT_SYMBOL_NAMESPACE and lets
it be defined in a more obvious part of the source file.
David
