From: Thomas Monjalon <thomas@monjalon.net>
To: Ferruh Yigit <ferruh.yigit@intel.com>
Cc: Bruce Richardson <bruce.richardson@intel.com>,
dev@dpdk.org, david.marchand@redhat.com
Subject: Re: [dpdk-dev] [PATCH v2] devtools: fix 32-bits build
Date: Mon, 09 Nov 2020 18:20:38 +0100 [thread overview]
Message-ID: <3526179.Ta2yc0qjKd@thomas> (raw)
In-Reply-To: <07886870-844e-ef22-912a-b945746e0c30@intel.com>
09/11/2020 18:15, Ferruh Yigit:
> On 11/9/2020 5:01 PM, Thomas Monjalon wrote:
> > 09/11/2020 17:48, Ferruh Yigit:
> >> On 11/9/2020 4:19 PM, Bruce Richardson wrote:
> >>> On Mon, Nov 09, 2020 at 05:14:24PM +0100, Thomas Monjalon wrote:
> >>>> 09/11/2020 16:44, Bruce Richardson:
> >>>>> On Mon, Nov 09, 2020 at 04:28:16PM +0100, Thomas Monjalon wrote:
> >>>>>> 09/11/2020 15:55, Ferruh Yigit:
> >>>>>>> If the 'PKG_CONFIG_PATH' is set in the environment before script run,
> >>>>>>> 32 bit still uses that value for 64-bits libraries.
> >>>>>>>
> >>>>>>> From the build log:
> >>>>>>>
> >>>>>>> "
> >>>>>>> Using DPDK_TARGET i386-pc-linux-gnu
> >>>>>>> meson -Dexamples=l3fwd --buildtype=debugoptimized --werror
> >>>>>>> -Dc_args=-m32 -Dc_link_args=-m32
> >>>>>>> /tmp/dpdk_maintain/self/dpdk/devtools/.. ./build-32b
> >>>>>>> ...
> >>>>>>> Using 'PKG_CONFIG_PATH' from environment with value:
> >>>>>>> '/usr/local/lib64/pkgconfig/
> >>>>>>> "
> >>>>>>>
> >>>>>>> This causes build error when linking with the found libraries.
> >>>>>>>
> >>>>>>> Reproduced with 'librte_bpf' which only has 64 bit installed but still
> >>>>>>> enables building 'af_xdp' and link fails.
> >>>>>>>
> >>>>>>> To fix this, using default 'PKG_CONFIG_PATH' variable unless
> >>>>>>> 'CUSTOM_PKG_CONFIG_PATH' set, and set the 'CUSTOM_PKG_CONFIG_PATH' for
> >>>>>>> 32 bit build.
> >>>>>>>
> >>>>>>> Fixes: 9b83106d8784 ("devtools: test 32-bit build")
> >>>>>>>
> >>>>>>> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
> >>>>>>> ---
> >>>>>>> Cc: Bruce Richardson <bruce.richardson@intel.com>
> >>>>>>> Cc: Thomas Monjalon <thomas@monjalon.net>
> >>>>>>>
> >>>>>>> v2:
> >>>>>>> * Enable overwriting default 'PKG_CONFIG_PATH' value
> >>>>>>
> >>>>>> It was not my conclusion.
> >>>>>> I think we can just reset all env vars.
> >>>>>>
> >>>>>
> >>>>> Perhaps, but I think that may cause more problems for people who want
> >>>>> custom CFLAGS and LDFLAGS for building with extra 3rd-party packages.
> >>>>> While this can be done using the config script that is sourced in, I
> >>>>> suspect most of us do not use such a script.
> >>>>>
> >>>>> Therefore I'd suggest rather than clearing the env vars globally, we just
> >>>>> override them temporarily for the 32-bit, using CFLAGS32, LDFLAGS32,
> >>>>> PKG_CONFIG_PATH32 instead. That allows someone to have a 32-bit and 64-bit
> >>>>> version of e.g. libbpf installed, with a PKG_CONFIG_PATH pointing to each.
> >>>>
> >>>> So you want to duplicate all vars just to avoid writing them cleanly in a file?
> >>>> And what happens for other targets?
> >>>>
> >>>> I'm tending to nack this approach.
> >>>> There is no problem in using a clean config file.
> >>>>
> >>> Well, when you put it that way :-)
> >>> However, I was only thinking of the 3 variables above, though really it's
> >>> only the PKG_CONFIG_PATH that is going to be the most likely culprit, so
> >>> supporting just one extra var "PKG_CONFIG_PATH_32" seems reasonable.
> >>>
> >>> I'm also ok with your solution of just clearing the environment and relying
> >>> on a config file too, though.
> >>>
> >>
> >> I was worried for any side affects to reset the var, this is smaller change to
> >> change only 'CUSTOM_PKG_CONFIG_PATH' for 'build-32b'
> >>
> >> I can send a new version to reset 'CUSTOM_PKG_CONFIG_PATH',
> >> what is the intention/plan with the config file?
> >
> > You should use a config file to set PKG_CONFIG_PATH, PATH,
> > DPDK_MESON_OPTIONS, etc accordingly to build with all
> > your non-installed dependencies.
> > CUSTOM_PKG_CONFIG_PATH is a way to avoid using the config file
> > in a single case.
> >
>
> Either use 'PKG_CONFIG_PATH' environment variable or a config file, both has the
> same problem that we need two different pkg config path, one for 64bit and one
> for 32bit.
No. In the config file, you can set the right value by testing $DPDK_TARGET.
Note: the config file is a shell script.
> v2 of this patch provides a way to use correct one when needed.
This is not a global fix.
> Moving the default values from system environment or script hardcoded values to
> a config file is something else and can be done later, what do you think?
The root cause of your issue is using variables from the environment.
The global fix is then to reset them all,
while keeping a default value for PATH which is not only for compilation config.
> > Don't be afraid breaking the assumption of theoritical users
> > of this script. If we want to avoid PATH reset,
> > we can add such line at the beginning of the script:
> > default_path=${PATH:-/bin:/usr/bin}
next prev parent reply other threads:[~2020-11-09 17:20 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-11-09 13:00 [dpdk-dev] [PATCH] " Ferruh Yigit
2020-11-09 13:19 ` Thomas Monjalon
2020-11-09 13:24 ` Ferruh Yigit
2020-11-09 13:35 ` Bruce Richardson
2020-11-09 13:51 ` Thomas Monjalon
2020-11-09 14:55 ` [dpdk-dev] [PATCH v2] " Ferruh Yigit
2020-11-09 15:28 ` Thomas Monjalon
2020-11-09 15:44 ` Bruce Richardson
2020-11-09 16:14 ` Thomas Monjalon
2020-11-09 16:19 ` Bruce Richardson
2020-11-09 16:48 ` Ferruh Yigit
2020-11-09 17:01 ` Thomas Monjalon
2020-11-09 17:15 ` Ferruh Yigit
2020-11-09 17:20 ` Thomas Monjalon [this message]
2020-11-09 17:44 ` Thomas Monjalon
2020-11-09 21:02 ` Thomas Monjalon
2020-11-11 12:34 ` Ferruh Yigit
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=3526179.Ta2yc0qjKd@thomas \
--to=thomas@monjalon.net \
--cc=bruce.richardson@intel.com \
--cc=david.marchand@redhat.com \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).