Hi,
Niko Tyni wrote:
> Looks like dpkg tries to sync /usr/bin/perl5.12.4.dpkg-new when a hard
> link has already been renamed to /usr/bin/perl and presumably gotten
> executed by someone else.
Ah, fun. :)
Does this help?
---
debian/changelog | 6 ++++++
src/archives.c | 4 ++--
2 files changed, 8 insertions(+), 2 deletions(-)
diff --git a/debian/changelog b/debian/changelog
index 20c296d5..5bab74e7 100644
--- a/debian/changelog
+++ b/debian/changelog
@@ -186,6 +186,12 @@ dpkg (1.16.1~jrn) local; urgency=low
* Add new --raw-extract option to dpkg-deb combining --control and
--extract. Closes: #552123
+ [ Jonathan Nieder ]
+ * Open extracted files for reading, not writing, in order to fsync() them.
+ Otherwise the open can error out when preparing to rename a binary into
+ place that has a hard link already in use. Regression introduced in
+ 1.15.6.1. Closes: #635683
+
[ Updated dpkg translations ]
* German (Sven Joachim). Closes: #620312
* Swedish (Peter Krefting).
diff --git a/src/archives.c b/src/archives.c
index f060e706..d0bb2d8f 100644
--- a/src/archives.c
+++ b/src/archives.c
@@ -894,7 +894,7 @@ tar_writeback_barrier(struct fileinlist *files, struct
pkginfo *pkg)
setupfnamevbs(usename);
- fd = open(fnamenewvb.buf, O_WRONLY);
+ fd = open(fnamenewvb.buf, O_RDONLY);
if (fd < 0)
ohshite(_("unable to open '%.255s'"), fnamenewvb.buf);
/* Ignore the return code as it should be considered equivalent to an
@@ -937,7 +937,7 @@ tar_deferred_extract(struct fileinlist *files, struct
pkginfo *pkg)
debug(dbg_eachfiledetail, "deferred extract needs fsync");
- fd = open(fnamenewvb.buf, O_WRONLY);
+ fd = open(fnamenewvb.buf, O_RDONLY);
if (fd < 0)
ohshite(_("unable to open '%.255s'"), fnamenewvb.buf);
if (fsync(fd))
--
1.7.6
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
