Bug#1016560: glibc 2.34 breaks scalpel autopkgtest on amd64: bash: line 1: 1961 Segmentation fault
control: tag -1 +pending
On 2022-08-09 09:31, Aurelien Jarno wrote:
> control: severity 1016560 serious
>
> On 2022-08-03 00:01, Aurelien Jarno wrote:
> > Source: scalpel
> > Version: 1.60-9
> > Severity: important
> > Tags: upstream patch
> > User: debian-gl...@lists.debian.org
> > Usertags: glibc2.34
> >
> > Dear maintainer,
> >
> > The autopkgtest of scalpel fails in sid on amd64 when that autopkgtest is
> > run with the binary packages of glibc from experimental. It passes when
> > run with only packages from sid. In tabular form:
> >
> > passfail
> > glibcfrom sid2.34-0experimental5
> > scalpel from sid1.60-9
> > all others from sidfrom sid
> >
> > Here is the relevant part of the test log:
> >
> > autopkgtest [10:36:40]: test command1: scalpel -c debian/tests/scalpel.conf
> > debian/tests/lua.img
> > autopkgtest [10:36:40]: test command1: [---
> >
> > Opening target
> > "/tmp/autopkgtest-lxc.93yq46zi/downtmp/build.fXk/src/debian/tests/lua.img"
> >
> > bash: line 1: 1961 Segmentation fault bash -ec 'scalpel -c
> > debian/tests/scalpel.conf debian/tests/lua.img' 2> >(tee -a
> > /tmp/autopkgtest-lxc.93yq46zi/downtmp/command1-stderr >&2) > >(tee -a
> > /tmp/autopkgtest-lxc.93yq46zi/downtmp/command1-stdout)
> >
> > The full test log is available there:
> > https://ci.debian.net/data/autopkgtest/unstable/amd64/s/scalpel/24235565/log.gz
> >
> > After some debugging, I have found the issue to be a duplicate use of a
> > va_list without using va_copy. Please find attached a patch to fix that.
> >
> > Regards
> > Aurelien
>
> > --- scalpel-1.60.orig/helpers.c
> > +++ scalpel-1.60/helpers.c
> > @@ -70,12 +70,14 @@ void setProgramName(char *s) {
> > // write entry to both the screen and the audit file
> > void scalpelLog(struct scalpelState *state, char *format, ...) {
> >
> > - va_list argp;
> > + va_list argp, argp2;
> >
> >va_start(argp,format);
> > + va_copy(argp2, argp);
> >vfprintf (stderr,format,argp);
> > - vfprintf (state->auditFile,format,argp);
> >va_end(argp);
> > + vfprintf (state->auditFile,format,argp2);
> > + va_end(argp2);
> > }
> >
> > // determine if two characters match, with optional case
>
> glibc 2.34 is now in unstable, upgrading the severity.
I have uploaded a NMU fixing this issue to DELAYED/2. Please find the
corresponding debdiff attached. Also please feel free to ask me to delay
or cancel this NMU.
Regards
Aurelien
--
Aurelien Jarno GPG: 4096R/1DDD8C9B
aurel...@aurel32.net http://www.aurel32.net
diff -Nru scalpel-1.60/debian/changelog scalpel-1.60/debian/changelog
--- scalpel-1.60/debian/changelog 2020-04-21 02:01:23.0 +
+++ scalpel-1.60/debian/changelog 2022-08-11 13:45:53.0 +
@@ -1,3 +1,11 @@
+scalpel (1.60-9+nmu1) unstable; urgency=medium
+
+ * Non-maintainer upload.
+ * debian/patches/60_va_copy.patch: fix the va_list usage in
+setProgramName(). (Closes: #1016560).
+
+ -- Aurelien Jarno Thu, 11 Aug 2022 15:45:53 +0200
+
scalpel (1.60-9) unstable; urgency=medium
* Team upload.
diff -Nru scalpel-1.60/debian/patches/60_va_copy.patch
scalpel-1.60/debian/patches/60_va_copy.patch
--- scalpel-1.60/debian/patches/60_va_copy.patch1970-01-01
00:00:00.0 +
+++ scalpel-1.60/debian/patches/60_va_copy.patch2022-08-11
13:44:57.0 +
@@ -0,0 +1,23 @@
+Description: Fix the va_list usage in setProgramName(). (Closes: #1016560).
+Author: Aurelien Jarno
+Last-Update: 2022-08-11
+
+--- scalpel-1.60.orig/helpers.c
scalpel-1.60/helpers.c
+@@ -70,12 +70,14 @@ void setProgramName(char *s) {
+ // write entry to both the screen and the audit file
+ void scalpelLog(struct scalpelState *state, char *format, ...) {
+
+- va_list argp;
++ va_list argp, argp2;
+
+ va_start(argp,format);
++ va_copy(argp2, argp);
+ vfprintf (stderr,format,argp);
+- vfprintf (state->auditFile,format,argp);
+ va_end(argp);
++ vfprintf (state->auditFile,format,argp2);
++ va_end(argp2);
+ }
+
+ // determine if two characters match, with optional case
diff -Nru scalpel-1.60/debian/patches/series scalpel-1.60/debian/patches/series
--- scalpel-1.60/debian/patches/series 2020-04-21 01:59:40.0 +
+++ scalpel-1.60/debian/patches/series 2022-08-11 13:45:53.0 +
@@ -3,3 +3,4 @@
30_improve-error-message.patch
40_fix-spelling.patch
50_fix-jpg-remove-spaces.patch
+60_va_copy.patch
Bug#1016560: glibc 2.34 breaks scalpel autopkgtest on amd64: bash: line 1: 1961 Segmentation fault
control: severity 1016560 serious
On 2022-08-03 00:01, Aurelien Jarno wrote:
> Source: scalpel
> Version: 1.60-9
> Severity: important
> Tags: upstream patch
> User: debian-gl...@lists.debian.org
> Usertags: glibc2.34
>
> Dear maintainer,
>
> The autopkgtest of scalpel fails in sid on amd64 when that autopkgtest is
> run with the binary packages of glibc from experimental. It passes when
> run with only packages from sid. In tabular form:
>
> passfail
> glibcfrom sid2.34-0experimental5
> scalpel from sid1.60-9
> all others from sidfrom sid
>
> Here is the relevant part of the test log:
>
> autopkgtest [10:36:40]: test command1: scalpel -c debian/tests/scalpel.conf
> debian/tests/lua.img
> autopkgtest [10:36:40]: test command1: [---
>
> Opening target
> "/tmp/autopkgtest-lxc.93yq46zi/downtmp/build.fXk/src/debian/tests/lua.img"
>
> bash: line 1: 1961 Segmentation fault bash -ec 'scalpel -c
> debian/tests/scalpel.conf debian/tests/lua.img' 2> >(tee -a
> /tmp/autopkgtest-lxc.93yq46zi/downtmp/command1-stderr >&2) > >(tee -a
> /tmp/autopkgtest-lxc.93yq46zi/downtmp/command1-stdout)
>
> The full test log is available there:
> https://ci.debian.net/data/autopkgtest/unstable/amd64/s/scalpel/24235565/log.gz
>
> After some debugging, I have found the issue to be a duplicate use of a
> va_list without using va_copy. Please find attached a patch to fix that.
>
> Regards
> Aurelien
> --- scalpel-1.60.orig/helpers.c
> +++ scalpel-1.60/helpers.c
> @@ -70,12 +70,14 @@ void setProgramName(char *s) {
> // write entry to both the screen and the audit file
> void scalpelLog(struct scalpelState *state, char *format, ...) {
>
> - va_list argp;
> + va_list argp, argp2;
>
>va_start(argp,format);
> + va_copy(argp2, argp);
>vfprintf (stderr,format,argp);
> - vfprintf (state->auditFile,format,argp);
>va_end(argp);
> + vfprintf (state->auditFile,format,argp2);
> + va_end(argp2);
> }
>
> // determine if two characters match, with optional case
glibc 2.34 is now in unstable, upgrading the severity.
--
Aurelien Jarno GPG: 4096R/1DDD8C9B
aurel...@aurel32.net http://www.aurel32.net
Bug#1016560: glibc 2.34 breaks scalpel autopkgtest on amd64: bash: line 1: 1961 Segmentation fault
Source: scalpel
Version: 1.60-9
Severity: important
Tags: upstream patch
User: debian-gl...@lists.debian.org
Usertags: glibc2.34
Dear maintainer,
The autopkgtest of scalpel fails in sid on amd64 when that autopkgtest is
run with the binary packages of glibc from experimental. It passes when
run with only packages from sid. In tabular form:
passfail
glibcfrom sid2.34-0experimental5
scalpel from sid1.60-9
all others from sidfrom sid
Here is the relevant part of the test log:
autopkgtest [10:36:40]: test command1: scalpel -c debian/tests/scalpel.conf
debian/tests/lua.img
autopkgtest [10:36:40]: test command1: [---
Opening target
"/tmp/autopkgtest-lxc.93yq46zi/downtmp/build.fXk/src/debian/tests/lua.img"
bash: line 1: 1961 Segmentation fault bash -ec 'scalpel -c
debian/tests/scalpel.conf debian/tests/lua.img' 2> >(tee -a
/tmp/autopkgtest-lxc.93yq46zi/downtmp/command1-stderr >&2) > >(tee -a
/tmp/autopkgtest-lxc.93yq46zi/downtmp/command1-stdout)
The full test log is available there:
https://ci.debian.net/data/autopkgtest/unstable/amd64/s/scalpel/24235565/log.gz
After some debugging, I have found the issue to be a duplicate use of a
va_list without using va_copy. Please find attached a patch to fix that.
Regards
Aurelien
--- scalpel-1.60.orig/helpers.c
+++ scalpel-1.60/helpers.c
@@ -70,12 +70,14 @@ void setProgramName(char *s) {
// write entry to both the screen and the audit file
void scalpelLog(struct scalpelState *state, char *format, ...) {
- va_list argp;
+ va_list argp, argp2;
va_start(argp,format);
+ va_copy(argp2, argp);
vfprintf (stderr,format,argp);
- vfprintf (state->auditFile,format,argp);
va_end(argp);
+ vfprintf (state->auditFile,format,argp2);
+ va_end(argp2);
}
// determine if two characters match, with optional case
