[dpdk-dev] [EXT] Re: checkpatch failing on map change

[Liron Himi]

Great. Thanks. Will send it soon

Regards,
Liron

-----Original Message-----
From: David Marchand <david.marchand at redhat.com> 
Sent: Tuesday, 17 December 2019 10:15
To: Liron Himi <lironh at marvell.com>; Cristian Dumitrescu <cristian.dumitrescu at intel.com>
Cc: dpdk-dev <dev at dpdk.org>; Thomas Monjalon <thomas at monjalon.net>; nhorman at tuxdriver.com; stephen at networkplumber.org
Subject: [EXT] Re: [dpdk-dev] checkpatch failing on map change

External Email

----------------------------------------------------------------------
Hello,

On Tue, Dec 17, 2019 at 8:36 AM Liron Himi <lironh at marvell.com> wrote:
> One of our drivers is using the rte_cfgfile and meson is failing on 'rte_cfgfile_section_num_entries_by_index' which is undefined.
>
> After a few debug cycles, I found out that this function is missing from the .map file.
>
> So I added it and created a patch but  checkpatch is failing with
> ERROR: symbol rte_cfgfile_section_num_entries_by_index is added in the 
> DPDK_20.0 section, but is expected to be added in the EXPERIMENTAL 
> section of the version map
>
> Do you know what it means and how to fix it? The function is not 
> defined as experimental so I don't understand what is the problem

Liron,

The script can't tell that the symbol that was added in the map file is already part of the stable ABI: so it complains that a new symbol should first be marked experimental.
But in this specific case, we can ignore the check.
You can send your patch to amend the map file.


Cristian,

Can you double check this library symbols to catch other unexported symbols?
A quick grep shows that rte_cfgfile_section_num_entries_by_index() is not called in the current tree, we have no UT on it.
Maybe something to add.


Thanks.

--
David Marchand