lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20240908102247.GA1175@1wt.eu>
Date: Sun, 8 Sep 2024 12:22:47 +0200
From: Willy Tarreau <w@....eu>
To: Shuah Khan <skhan@...uxfoundation.org>
Cc: Thomas Weißschuh <linux@...ssschuh.net>,
        "Paul E. McKenney" <paulmck@...nel.org>, linux-kernel@...r.kernel.org
Subject: Re: [GIT PULL] nolibc for 6.12-rc1

On Thu, Sep 05, 2024 at 05:57:52PM -0600, Shuah Khan wrote:
> On 9/5/24 10:35, Willy Tarreau wrote:
> > On Thu, Sep 05, 2024 at 05:57:22PM +0200, Thomas Weißschuh wrote:
> > > On 2024-09-05 08:32:14+0000, Willy Tarreau wrote:
> > > > On Thu, Sep 05, 2024 at 08:22:18AM +0200, Willy Tarreau wrote:
> > > > > > 
> > > > > > ./run-tests.sh -p -m user
> > > > > > 
> > > > > > These toolchains can then also be used for direct "make" invocations
> > > > > > through CROSS_COMPILE.
> > > > > 
> > > > > I really suspect an empty CC variable somewhere that could explain why
> > > > > only CROSS_COMPILE is used. I'll try to find time today to give it a
> > > > > try here as well, just in case I can reproduce the same issue.
> > > > 
> > > > In fact I'm getting it without any options:
> > > > 
> > > >    $ ./run-tests.sh
> > > >    realpath: /home/willy/.cache/crosstools/gcc-13.2.0-nolibc/i386-linux/bin/i386-linux-: No such file or directory
> > > > 
> > > > It comes from here in test_arch():
> > > > 
> > > >          cross_compile=$(realpath "${download_location}gcc-${crosstool_version}-nolibc/${ct_arch}-${ct_abi}/bin/${ct_arch}-${ct_abi}-")
> > > > 
> > > > Thus it's indeed related to the absence of the toolchain there. It's
> > > > just that the way the error is reported (due to set -e) is a bit harsh.
> > > 
> > > Ack. It should not occur with "-p" though.
> > 
> > Agreed, I was focusing on first experience for users essentially.
> > 
> > > > What about this ?
> > > > 
> > > >    $ ./run-tests.sh
> > > >    No toolchain found in /home/willy/.cache/crosstools/gcc-13.2.0-nolibc/i386-linux.
> > > >    Did you install the toolchains or set the correct arch ? Rerun with -h for help.
> > > >    Aborting...
> > > > 
> > > > or anything similar, achieved by this patch (warning copy-paste, mangled
> > > > indents):
> > > > 
> > > > diff --git a/tools/testing/selftests/nolibc/run-tests.sh b/tools/testing/selftests/nolibc/run-tests.sh
> > > > index e7ecda4ae796..0f67e80051dc 100755
> > > > --- a/tools/testing/selftests/nolibc/run-tests.sh
> > > > +++ b/tools/testing/selftests/nolibc/run-tests.sh
> > > > @@ -143,6 +143,13 @@ test_arch() {
> > > >          arch=$1
> > > >          ct_arch=$(crosstool_arch "$arch")
> > > >          ct_abi=$(crosstool_abi "$1")
> > > > +
> > > > +       if [ ! -d "${download_location}gcc-${crosstool_version}-nolibc/${ct_arch}-${ct_abi}/bin/." ]; then
> > > > +               echo "No toolchain found in ${download_location}gcc-${crosstool_version}-nolibc/${ct_arch}-${ct_abi}."
> > > > +               echo "Did you install the toolchains or set the correct arch ? Rerun with -h for help."
> > > > +               return 1
> > > > +       fi
> > > > +
> > > >          cross_compile=$(realpath "${download_location}gcc-${crosstool_version}-nolibc/${ct_arch}-${ct_abi}/bin/${ct_arch}-${ct_abi}-")
> > > >          build_dir="${build_location}/${arch}"
> > > >          if [ "$werror" -ne 0 ]; then
> > > 
> > > Looks good.
> > 
> > OK thanks, I'll try to handle it this week-end if I'm not beaten to
> > it. If you or Shuah want to merge it before, feel free to, no offense
> > on my side!
> > 
> 
> Sounds good. My system is back to a good state with the tests after running
> ./run-tests.sh -p -m user
> 
> My guess is my setup was lost when I upgraded my system.

Makes sense.

FWIW I've just pushed the patch above to nolibc-next. Since it's not a
result of a change of the last PR, there's no need to udpate it I think,
it can easily wait for the next one (it's just a help message after all).

During the tests I've also got caught with -d which needs a trailing slash
otherwise doesn't concatenate directories correctly. I don't know if that's
intentional or not so I didn't change it (and it's not important either).

Best regards,
Willy

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ