summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorGuillem Jover <guillem@debian.org>2019-01-23 12:15:26 +0100
committerGuillem Jover <guillem@debian.org>2019-01-23 12:16:16 +0100
commitc1e62efcf183e9d741d36390b43e1d07991e4979 (patch)
tree21b08febccf88cf7fd53ed7ad63728d3bd439ff9
parent15c2161f7d8093b2ec857c343987339f59c8afd5 (diff)
downloaddpkg-c1e62efcf183e9d741d36390b43e1d07991e4979.tar.gz
s-s-d: Remove bogus comment about Linux user namespaces vs /dev/null
The /dev/null device always has world-writable permissions, so it would always fail anyway. Spotted-by: James Clarke <jrtc27@debian.org> (on IRC)
-rw-r--r--debian/changelog3
-rw-r--r--utils/start-stop-daemon.c3
2 files changed, 2 insertions, 4 deletions
diff --git a/debian/changelog b/debian/changelog
index f397b2642..1b0c6a33b 100644
--- a/debian/changelog
+++ b/debian/changelog
@@ -2,8 +2,7 @@ dpkg (1.19.4) UNRELEASED; urgency=medium
* start-stop-daemon: Do not sanity check the pidfile when it is specified as
/dev/null, as that implies the caller wants to start the program no matter
- what. Even though the check should not fail on this condition, it does on
- some scenarios, such as when using Linux user namespaces. Closes: #920242
+ what. Closes: #920242
* Portability:
- start-stop-daemon: Only use SO_PASSCRED if defined. Fixes build failure
at least on GNU/Hurd.
diff --git a/utils/start-stop-daemon.c b/utils/start-stop-daemon.c
index f01a8982d..7a52643f8 100644
--- a/utils/start-stop-daemon.c
+++ b/utils/start-stop-daemon.c
@@ -2269,8 +2269,7 @@ do_pidfile(const char *name)
*
* If we got /dev/null specified as the pidfile, we ignore the
* checks, as this is being used to run processes no matter
- * what. Even though the checks should not fail, they do on
- * some scenarios, such as when using Linux user namespaces. */
+ * what. */
if (match_mode == MATCH_PIDFILE &&
strcmp(name, "/dev/null") != 0) {
struct stat st;