cobra output cygwin vs linux

[50AmpFuse]

When running cobra under Win10 (cygwin), e.g. cobra -f misra2012 test.c, the first two lines of output are always:

trailing backslash (\)
empty (sub)expression

My guess is this output is generated by one of the external tools cobra is calling, i.e. grep, sed, find, etc. In linux (Ubuntu 22.04.5 LTS) i don't see this behavior when running cobra with the same parameters on the same code.

cygwin: grep --version

grep (GNU grep) 3.12
Packaged by Cygwin (3.12-1)
Copyright (C) 2025 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <https://gnu.org/licenses/gpl.html>.
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.

Written by Mike Haertel and others; see
<https://git.savannah.gnu.org/cgit/grep.git/tree/AUTHORS>.

grep -P uses PCRE2 10.45 2025-02-05

cygwin: sed --version

sed (GNU sed) 4.9
Packaged by Cygwin (4.9-1)
Copyright (C) 2022 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <https://gnu.org/licenses/gpl.html>.
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.

Written by Jay Fenlason, Tom Lord, Ken Pizzini,
Paolo Bonzini, Jim Meyering, and Assaf Gordon.

This sed program was built without SELinux support.

GNU sed home page: <https://www.gnu.org/software/sed/>.
General help using GNU software: <https://www.gnu.org/gethelp/>.
E-mail bug reports to: <bug-sed@gnu.org>.

cygwin: find --version

find (GNU findutils) 4.10.0
Packaged by Cygwin (4.10.0-1)
Copyright (C) 2024 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <https://gnu.org/licenses/gpl.html>.
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.

Written by Eric B. Decker, James Youngman, and Kevin Dalley.
Features enabled: D_TYPE O_NOFOLLOW(enabled) LEAF_OPTIMISATION FTS(FTS_CWDFD) CBO(level=1)

[nimble-code]

this does not happen on my system (Windows 11 with cygwin) not sure what can cause this on Win10/cygwin?

…

On Fri, Aug 29, 2025 at 6:56 AM Sebastian ***@***.***> wrote: *50AmpFuse* created an issue (nimble-code/Cobra#83) <#83> When running cobra under Win10 (cygwin), e.g. cobra -f misra2012 test.c, the first two lines of output are always: trailing backslash (\) empty (sub)expression My guess is this output is generated by one of the external tools cobra is calling, i.e. grep, sed, find, etc. In linux (Ubuntu 22.04.5 LTS) i don't see this behavior when running cobra with the same parameters on the same code. — Reply to this email directly, view it on GitHub <#83>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AK6L6IIRQJB7EYLPMUA7JKT3QBLZZAVCNFSM6AAAAACFEUQW7OVHI2DSMVQWIX3LMV43ASLTON2WKOZTGM3DMOJWGQ4DQOI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[50AmpFuse]

Addition: this applies only to the misra2012 rule check, not the other checks, even not misra1997 and misra2004.

Only the P10 rule check has a similar issue, it always gives this line as first output:

empty (sub)expression

Again only on Win10/cygwin, not on linux. I'll check on Win11 in near future.

[50AmpFuse]

I traced down the issue to cobra_lib.c, function regstart(), specifically this part (lines 161 - 166):

n = regcomp(&rex[p], s, REG_NOSUB|REG_EXTENDED);
if (n != 0)	// compile
{	regerror(n, &rex[p], ebuf, sizeof(ebuf));
	printf("%s\n", ebuf);
	return 0;
}

On cygwin regcomp() returns non zero for s =

^['"]\x[0-9][0-9][^'"]

resulting in regerror() message

trailing backslash (\)

and for s = "" (strlen() = 0) resulting in regerror() message

empty (sub)expression

On Ubuntu these messages don't appear, i.e. regcomp() returns zero in these cases. So somehow the behavior of regcomp() is different in cygwin compared to linux.

On Win11 it showes the same behavior as Win10.

[nimble-code]

this is odd -- I put in a printf statement of n and s just before the test for zero in regstart (right after regcomp) n = regcomp(&rex[p], s, REG_NOSUB|REG_EXTENDED); printf("n=%d '%s'\n", n, s); and I get this result when I feed it that string you gave: n=0 '^['"]\x[0-9][0-9][^'"]' so, n==0, not non-zero on my system -g

…

On Tue, Sep 16, 2025 at 7:51 AM Sebastian ***@***.***> wrote: *50AmpFuse* left a comment (nimble-code/Cobra#83) <#83 (comment)> I traced down the issue to cobra_lib.c, function regstart(), specifically this part (lines 161 - 166): n = regcomp(&rex[p], s, REG_NOSUB|REG_EXTENDED);if (n != 0) // compile { regerror(n, &rex[p], ebuf, sizeof(ebuf)); printf("%s\n", ebuf); return 0; } On cygwin regcomp() returns non zero for s = ^['"]\x[0-9][0-9][^'"] resulting in regerror() message trailing backslash (\) and for s = "" (strlen() = 0) resulting in regerror() message empty (sub)expression On Ubuntu these messages don't appear, i.e. regcomp() returns zero. On Win11 it showes the same behavior as Win10. — Reply to this email directly, view it on GitHub <#83 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AK6L6IM3CZZGEZIGXWVKG6T3TAPVVAVCNFSM6AAAAACFEUQW7OVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTEOJZGEZTIMRWGQ> . You are receiving this because you commented.Message ID: ***@***.***>

[nimble-code]

(this is on cygwin64 on Windows 11) On Tue, Sep 16, 2025 at 2:15 PM Gerard Holzmann ***@***.***> wrote:

…

this is odd -- I put in a printf statement of n and s just before the test for zero in regstart (right after regcomp) n = regcomp(&rex[p], s, REG_NOSUB|REG_EXTENDED); printf("n=%d '%s'\n", n, s); and I get this result when I feed it that string you gave: n=0 '^['"]\x[0-9][0-9][^'"]' so, n==0, not non-zero on my system -g On Tue, Sep 16, 2025 at 7:51 AM Sebastian ***@***.***> wrote: > *50AmpFuse* left a comment (nimble-code/Cobra#83) > <#83 (comment)> > > I traced down the issue to cobra_lib.c, function regstart(), specifically > this part (lines 161 - 166): > > n = regcomp(&rex[p], s, REG_NOSUB|REG_EXTENDED);if (n != 0) // compile > { regerror(n, &rex[p], ebuf, sizeof(ebuf)); > printf("%s\n", ebuf); > return 0; > } > > On cygwin regcomp() returns non zero for s = > > ^['"]\x[0-9][0-9][^'"] > > resulting in regerror() message > > trailing backslash (\) > > and for s = "" (strlen() = 0) resulting in regerror() message > > empty (sub)expression > > On Ubuntu these messages don't appear, i.e. regcomp() returns zero. > On Win11 it showes the same behavior as Win10. > > — > Reply to this email directly, view it on GitHub > <#83 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AK6L6IM3CZZGEZIGXWVKG6T3TAPVVAVCNFSM6AAAAACFEUQW7OVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTEOJZGEZTIMRWGQ> > . > You are receiving this because you commented.Message ID: > ***@***.***> >

[50AmpFuse]

It seems that the cygwin version plays a role in this issue. I was using a fairly recent version (cygwin1.dll 3.6.4) which gave me this issue. Now i've installed an older version (cygwin1.dll 3.3.3 from 2021) and the problem is gone!

This may be related, in 2023 cygwin made changes in the regex code:
https://sourceware.org/pipermail/cygwin-cvs/2023q1/015908.html

[50AmpFuse]

i.e. the behavior of cobra with cygwin1.dll 3.3.3 is equal to the behavior on Ubuntu, no regcomp() error messages.

[nimble-code]

Interesting -- I just checked and I'm using cygwin1.dll version 3.4.7, which also doesn't have the issue good reason not to update it I guess!

…

On Thu, Sep 18, 2025 at 7:14 AM Sebastian ***@***.***> wrote: *50AmpFuse* left a comment (nimble-code/Cobra#83) <#83 (comment)> It seems that the cygwin version plays a role in this issue. I was using a fairly recent version (cygwin1.dll 3.6.4) which gave me this issue. Now i've installed an older version (cygwin1.dll 3.3.3 from 2021) and the problem is gone! This may be related, in 2023 cygwin made changes in the regex code: https://sourceware.org/pipermail/cygwin-cvs/2023q1/015908.html — Reply to this email directly, view it on GitHub <#83 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AK6L6IMCCYXRN3LVSFEKNFD3TK43BAVCNFSM6AAAAACFEUQW7OVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTGMBXG4YDANBTGY> . You are receiving this because you commented.Message ID: ***@***.***>

[50AmpFuse]

Well, i'm not sure.
Printing "empty (sub)expression" if regcomp() is called on an empty string isn't so bad?
I can't really judge the other case, i lack regex knowledge here.

And what's the impact on cobra results if regstart() returns 0 instead of 1...?

Further I would like to propose to change

n = regcomp(&rex[p], s, REG_NOSUB|REG_EXTENDED);
if (n != 0)	// compile
{	regerror(n, &rex[p], ebuf, sizeof(ebuf));
	printf("%s\n", ebuf);
	return 0;
}

into something like

n = regcomp(&rex[p], s, REG_NOSUB|REG_EXTENDED);
if (n != 0)	// compile
{	regerror(n, &rex[p], ebuf, sizeof(ebuf));
	printf("regcomp() on string '%s' returned: %s\n", s, ebuf);
	return 0;
}

to make the source of these messages clear.

[nimble-code]

good point -- i'll have that change in the next version

…

On Fri, Sep 19, 2025 at 3:18 AM Sebastian ***@***.***> wrote: *50AmpFuse* left a comment (nimble-code/Cobra#83) <#83 (comment)> Well, i'm not sure. Printing "empty (sub)expression" if regcomp() is called on an empty string isn't so bad? I can't really judge the other case, i lack regex knowledge here. And what's the impact on cobra results if regstart() returns 0 instead of 1...? Further I would like to propose to change n = regcomp(&rex[p], s, REG_NOSUB|REG_EXTENDED);if (n != 0) // compile { regerror(n, &rex[p], ebuf, sizeof(ebuf)); printf("%s\n", ebuf); return 0; } into something like n = regcomp(&rex[p], s, REG_NOSUB|REG_EXTENDED);if (n != 0) // compile { regerror(n, &rex[p], ebuf, sizeof(ebuf)); printf("regcomp() on string '%s' returned: %s\n", s, ebuf); return 0; } to make the source of these messages clear. — Reply to this email directly, view it on GitHub <#83 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AK6L6ILPYQUKFLKX66WJO4D3TPJ6PAVCNFSM6AAAAACFEUQW7OVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTGMJRGU4TKMRZGM> . You are receiving this because you commented.Message ID: ***@***.***>