Hi Bruce,
Sorry for the late reply. Holiday time in Israel.
Thursday, September 13, 2018 12:22 PM, Bruce Richardson:
> Subject: Re: [dpdk-dev] [PATCH v5] net/mlx: add meson build support
>
> On Fri, Sep 07, 2018 at 11:34:29AM +0100, Bruce Richardson wrote:
> > On Wed, Sep 05, 2018 at 02:47:46PM +0300, Shahaf Shuler wrote:
> > > From: Nelio Laranjeiro <nelio.laranje...@6wind.com>
[...]
> > couple of minor points:
> > * These new additions use tab rather than space for indent. The standard
> > with meson I think is normally spaces, but I (inadvertently) added all
> > other DPDK meson.build files with tabs, so for consistency I think it
> > would be good if the whole file used tabs i.e. change all the other lines
> > apart from your new ones in v5.
OK, all indentation will be with tab.
> >
> > * is the "rm" command really necessary? I would expect the
> configure_file()
> > function to just do the right thing here when replacing the file.
Yep it is not needed. Also from meson sources:
output = kwargs['output']
ofile_rpath = os.path.join(self.subdir, output)
if not isinstance(output, str):
raise InterpreterException('Output file name must be a string')
if ofile_rpath in self.configure_file_outputs:
mesonbuildfile = os.path.join(self.subdir, 'meson.build')
current_call = "{}:{}".format(mesonbuildfile, self.current_lineno)
first_call = "{}:{}".format(mesonbuildfile,
self.configure_file_outputs[ofile_rpath])
mlog.warning('Output file', mlog.bold(ofile_rpath, True), 'for
configure_file() at', current_call, 'o
verwrites configure_file() output at', first_call)
> >
> > * Would you consider splitting the args array into two arrays, one for the
> > "type" values and another for the enum/define ones? This would mean
> > having two loops, but the loops themselves would be clearer and shorter -
> > as would the arrays. Especially since according to
> >
> https://emea01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fmes
> onbuild.com%2FConfiguration.html&data=02%7C01%7Cshahafs%40mell
> anox.com%7C38f62333815c48eaea6e08d6195a908a%7Ca652971c7d2e4d9ba6
> a4d149256f461b%7C0%7C0%7C636724274091500993&sdata=fZvTg16D8K
> TsE7Q0DWsS9hiNnB%2BhHPuHi%2FFC7wrk8x8%3D&reserved=0
> > a boolean false converts to an "undef", rather than a 0/1 value.
> > For example:
> >
> > foreach arg:has_sym_args
> > config.set(arg[0], cc.has_header_symbol(arg[1], arg[3]))
> > endforeach
> >
Done.
> > * Very minor suggestion: is it neater to use "args: '-include ' + arg[1]",
> > instead of using prefix with the #include syntax? [Assuming the former
> > works]
I don't understand the benefit with this change. Is there?
> >
>
> Any plans for a new version of this? If so, can I also suggest splitting the
> patch
> into two, one for mlx4 and the other for mlx5.
No problem to split.
Another change in plans:
diff --git a/meson_options.txt b/meson_options.txt
index 484f3e2601..444a380d97 100644
--- a/meson_options.txt
+++ b/meson_options.txt
@@ -1,7 +1,7 @@
option('allow_invalid_socket_id', type: 'boolean', value: false,
description: 'allow out-of-range NUMA socket id\'s for platforms that
don\'t report the value correctly')
option('enable_driver_mlx_glue', type: 'boolean', value: false,
- description: 'Enable glue library for Mellanox ConnectX-3/4/5 NIC PMD')
+ description: 'Enable glue library for Mellanox PMDs')
option('enable_kmods', type: 'boolean', value: true,
description: 'build kernel modules')
option('examples', type: 'string', value: '',
I don't see the need to constantly update with each device being added.
>
> Regards,
> /Bruce
