6:33 a.m.
Hi Peter,
I love your patch! Yet something to improve:
[auto build test ERROR on linux/master]
[also build test ERROR on linus/master hnaz-linux-mm/master v5.12-rc2 next-20210311]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]
url:
https://github.com/0day-ci/linux/commits/Peter-Oberparleiter/gcov-fail-bu...
base: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
a74e6a014c9d4d4161061f770c9b4f98372ac778
config: nds32-randconfig-s032-20210311 (attached as .config)
compiler: nds32le-linux-gcc (GCC) 9.3.0
reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O
~/bin/make.cross
chmod +x ~/bin/make.cross
# apt-get install sparse
# sparse version: v0.6.3-262-g5e674421-dirty
#
https://github.com/0day-ci/linux/commit/01c5ab69a25e3cdfeff9cccf0f1fae26b...
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review
Peter-Oberparleiter/gcov-fail-build-on-gcov_info-size-mismatch/20210311-211208
git checkout 01c5ab69a25e3cdfeff9cccf0f1fae26b583cc39
# save the attached .config to linux build tree
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-9.3.0 make.cross C=1
CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' ARCH=nds32
If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp(a)intel.com>
All errors (new ones prefixed by >>):
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org
7:02 a.m.
New subject: [PATCH] gcov: fail build on gcov_info size mismatch
On Thu, Mar 11, 2021 at 11:34 AM kernel test robot <lkp(a)intel.com> wrote:
>> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
Wth? I'm not seeing how this can fail on nds32 - doesn't look like a
bashism, everything is properly quoted, etc etc. Plus it's a
cross-compile anyway, so the shell in question should be the same as
on all the other architectures.
Presumably the nds32 assembly contains something odd and unexpected,
but with the quoting, I can't see how even that could matter.
Yeah, the test itself could probably be simplified to testing both
conditions at the same time:
[ "$a $b" = ".size .LPBX0," ]
but that's a separate issue.
Funky. What am I missing? Presumably something really stupid.
Linus
2:49 p.m.
New subject: [PATCH] gcov: fail build on gcov_info size mismatch
On 3/12/21 4:02 AM, Linus Torvalds wrote:
On Thu, Mar 11, 2021 at 11:34 AM kernel test robot
<lkp(a)intel.com> wrote:
>>> kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
Wth? I'm not seeing how this can fail on nds32 - doesn't look like a
bashism, everything is properly quoted, etc etc. Plus it's a
cross-compile anyway, so the shell in question should be the same as
on all the other architectures.
Presumably the nds32 assembly contains something odd and unexpected,
but with the quoting, I can't see how even that could matter.
Yeah, the test itself could probably be simplified to testing both
conditions at the same time:
[ "$a $b" = ".size .LPBX0," ]
but that's a separate issue.
Funky. What am I missing? Presumably something really stupid.
Linus
Hi Linus,
The issue is from a=!, and [ "$a $b" = ".size .LPBX0," ] can avoid the
error.
+ [ ! = .size -a ABI = .LPBX0, ]
./kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
Best Regards,
Rong Chen
4:52 a.m.
New subject: [PATCH] gcov: fail build on gcov_info size mismatch
On Thu, Mar 11, 2021 at 7:50 PM Rong Chen <rong.a.chen(a)intel.com> wrote:
The issue is from a=!, and [ "$a $b" = ".size .LPBX0," ] can avoid
the
error.
+ [ ! = .size -a ABI = .LPBX0, ]
./kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
But that's not what the patch did.
The patch used quotes around $a, so "$a" should still be fine.
See:
[torvalds@ryzen ~]$ a="!" [ "$a" = ".size" ]
is fine, but
[torvalds@ryzen ~]$ a="!" [ $a = ".size" ]
-bash: [: =: unary operator expected
and the patch I saw, and that the test robot replied to, had that
correct quoting, afaik.
So I still don't see what the test robot is complaining about. Was
there an earlier version of the patch without the quotes that I didn't
see?
Or is the shell on the test robot doing something really really odd,
and it's somehow nds32-specific?
Linus
1:31 p.m.
New subject: [PATCH] gcov: fail build on gcov_info size mismatch
On 3/13/21 1:52 AM, Linus Torvalds wrote:
On Thu, Mar 11, 2021 at 7:50 PM Rong Chen
<rong.a.chen(a)intel.com> wrote:
>
> The issue is from a=!, and [ "$a $b" = ".size .LPBX0," ] can
avoid the
> error.
>
> + [ ! = .size -a ABI = .LPBX0, ]
> ./kernel/gcov/geninfosize.sh: 13: [: =: unexpected operator
But that's not what the patch did.
The patch used quotes around $a, so "$a" should still be fine.
See:
[torvalds@ryzen ~]$ a="!" [ "$a" = ".size" ]
is fine, but
[torvalds@ryzen ~]$ a="!" [ $a = ".size" ]
-bash: [: =: unary operator expected
and the patch I saw, and that the test robot replied to, had that
correct quoting, afaik.
So I still don't see what the test robot is complaining about. Was
there an earlier version of the patch without the quotes that I didn't
see?
Or is the shell on the test robot doing something really really odd,
and it's somehow nds32-specific?
Linus
Hi Linus,
It can be reproduced with '-a' option in dash:
$ a="!"
$ [ "$a" = ".size" ]
$ [ "$a" = ".size" -a "$b" = ".LPBX0," ]
sh: 2: [: =: unexpected operator
and there is a advice for the option at
https://wiki.ubuntu.com/DashAsBinSh, I'm not sure it's the best practice
or not.
While dash supports most uses of the -a and -o options, they have
very confusing semantics even in bash and are best avoided. Commands
like the following:
[ \( "$foo" = "$bar" -a -f /bin/baz \) -o ! -x /bin/quux ]
should be replaced with:
(([ "$foo" = "$bar" ] && [ -f /bin/baz ]) || [ ! -x
/bin/quux ])
Best Regards,
Rong Chen
6:12 a.m.
New subject: [PATCH] gcov: fail build on gcov_info size mismatch
On Sun, Mar 14, 2021 at 7:32 PM Rong Chen <rong.a.chen(a)intel.com> wrote:
It can be reproduced with '-a' option in dash:
Oh, ok. That kind of explains it.
'dash' is trash. Please somebody make a bug report.
$ a="!"
$ [ "$a" = ".size" ]
$ [ "$a" = ".size" -a "$b" = ".LPBX0," ]
sh: 2: [: =: unexpected operator
This is 100% a dash bug. There is no question what-so-ever about it.
This is not some kind of "POSIX is ambiguous", or "the handling of
'-a' is complicated".
It's simply just that dash is buggy.
While dash supports most uses of the -a and -o options, they have
very confusing semantics even in bash and are best avoided.
No, they have perfectly sane semantics in bash, and in POSIX.
See
https://pubs.opengroup.org/onlinepubs/9699919799/utilities/test.html
and there is absolutely zero question that bash does this correctly,
and dash does not.
But yes, it seems to be easy to work around, but still - could some
Ubuntu person please open a bug report on dash?
Linus
7:32 a.m.
New subject: [PATCH] gcov: fail build on gcov_info size mismatch
Linus Torvalds wrote:
On Sun, Mar 14, 2021 at 7:32 PM Rong Chen
<rong.a.chen(a)intel.com> wrote:
>
> It can be reproduced with '-a' option in dash:
Oh, ok. That kind of explains it.
'dash' is trash. Please somebody make a bug report.
> $ a="!"
> $ [ "$a" = ".size" ]
> $ [ "$a" = ".size" -a "$b" = ".LPBX0,"
]
> sh: 2: [: =: unexpected operator
This is 100% a dash bug. There is no question what-so-ever about it.
This is not some kind of "POSIX is ambiguous", or "the handling of
'-a' is complicated".
It's simply just that dash is buggy.
> While dash supports most uses of the -a and -o options, they have
> very confusing semantics even in bash and are best avoided.
No, they have perfectly sane semantics in bash, and in POSIX.
See
https://pubs.opengroup.org/onlinepubs/9699919799/utilities/test.html
and there is absolutely zero question that bash does this correctly,
and dash does not.
But yes, it seems to be easy to work around, but still - could some
Ubuntu person please open a bug report on dash?
fwiw, https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=850202
7:57 a.m.
New subject: [PATCH] gcov: fail build on gcov_info size mismatch
On Mon, Mar 15, 2021 at 1:32 PM Jamie Heilman
<jamie(a)audible.transient.net> wrote:
Yup, that seems to be the exact same thing from 4 years ago.
But it looks like nothing ever came out of it. It probably stayed
within the Debian bugzilla, and didn't go to upstream dash
maintainers.
It does look like dash is actually actively maintained, and it's even
a kernel maintainer that does it: Herbert Xu seems to maintain the
dash tree and I see commits from January.
So maybe we can get it fixed by just cc'ing Herbert.
Herbert, easy test-case:
$ [ "!" = ".size" ]
works, but
$ [ "!" = ".size" -a "b" = ".LPBX0," ]
causes
dash: 6: [: =: unexpected operator
because for some reason that "-a" ends up (wild handwaving here about
what is going on) re-parsing the first expression, and ignoring the
quoting around "!" when it does so.
I verified that the bug still exists in that current dash source tree,
but I didn't dig any deeper than that "wild handwaving guess" as to
what is actually going on.
Linus
10:09 a.m.
New subject: [PATCH] gcov: fail build on gcov_info size mismatch
On Mon, Mar 15, 2021 at 01:57:58PM -0700, Linus Torvalds wrote:
$ [ "!" = ".size" -a "b" = ".LPBX0," ]
causes
dash: 6: [: =: unexpected operator
because for some reason that "-a" ends up (wild handwaving here about
what is going on) re-parsing the first expression, and ignoring the
quoting around "!" when it does so.
The quoting on "!" doesn't help I'm afraid. Even though [ is a
built-in it is not allowed to look at the quoting because it's
supposed to behave in the same way whether you get the builtin [
or the one from /usr/bin.
So when it gets the expression the quoting on the "!" has already
been removed.
IOW this expression is ambiguous and may or may not fail depending
on how it's parsed.
Note that when you have a simple expression like
[ "!" = ".size" ]
special rules come into play on how it is parsed:
https://pubs.opengroup.org/onlinepubs/009604499/utilities/test.html
But this does not apply when you construct more complex ones with
-a.
There are two ways around this when writing scripts, you either
add something to ensure that "!" cannot occur, e.g.,
[ "X$a" = "X$b" -a ... ]
or you break it down into a simpler expression that is guaranteed
by POSIX:
[ "$a" = "$b" ] && [ ... ]
Cheers,
--
Email: Herbert Xu <herbert(a)gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
10:22 a.m.
New subject: [PATCH] gcov: fail build on gcov_info size mismatch
On Mon, Mar 15, 2021 at 4:10 PM Herbert Xu <herbert(a)gondor.apana.org.au> wrote:
The quoting on "!" doesn't help I'm afraid. Even though [ is a
built-in it is not allowed to look at the quoting because it's
supposed to behave in the same way whether you get the builtin [
or the one from /usr/bin.
So when it gets the expression the quoting on the "!" has already
been removed.
Oh Gods, that's truly disgusting.
But I guess from a historical perspective, it actually makes a nasty
kind of horrible sense.
What a crock. Oh well. We had the workaround already, so it's not like
this causes any heartache aside from "shell is even worse than most
people can imagine".
Linus
8:51 p.m.
New subject: [PATCH] gcov: fail build on gcov_info size mismatch
Linus Torvalds <torvalds(a)linux-foundation.org> wrote:
See:
[torvalds@ryzen ~]$ a="!" [ "$a" = ".size" ]
is fine, but
[torvalds@ryzen ~]$ a="!" [ $a = ".size" ]
-bash: [: =: unary operator expected
This isn't doing what you think it's doing. The first assignment
to a is not in effect when the shell is expanding $a so what you're
actually running is
a="!" [ = .size ]
Which is why it bombs out.
To get the desired result you need a semicolon:
$ a="!"; [ $a = ".size" ]
$
Cheers,
--
Email: Herbert Xu <herbert(a)gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
553
days inactive
558
days old
10 comments
5 participants
participants (5)
-
Herbert Xu -
Jamie Heilman -
kernel test robot -
Linus Torvalds -
Rong Chen