Re: ext2 - ext3 migration require reindexing?
On Sun, 2005-12-18 at 17:19 -0500, Ken Kennedy wrote: Anyway...quick question. I've started playing around with beagled, and realized that I probably should (finally) upsize my /home partition from ext2 to ext3. Should I/do I need to redo the indexing after that? If so, what should I whack? .beagle/Indexes, or just beagle/Indexes/FileSystemIndex? (ah, I see the FileAttributesStore.db in there now...this is starting to make sense). You should not need to redo any indexing, no. An up-conversion from ext2 to ext3 really does not entail anything but the creation of the journal. Your data and metadata are all fine. Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
Re: ANNOUNCE: Beagle 0.1.2
On Fri, 2005-11-11 at 14:33 -0500, Joe Shaw wrote: It is possible to find the k-th largest element of a unsorted list of N numbers in O(N) time: http://en.wikipedia.org/wiki/Selection_algorithm Thank goodness. I was worried you would not maintain Jon's standards. Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
Re: You need kernel 2.6.13
On Mon, 2005-08-01 at 22:49 +0300, Aviram Jenik wrote: Hi, Running beagled (CVS version) produces the following error: inotify_init: Function not implemented Inotify not supported! You need a 2.6.13 kernel or later with CONFIG_INOTIFY I'm running kernel 2.6.12.2 with the inotify-0.23-rml-2.6.12-15.patch and CONFIG_INOTIFY=y. I don't think I can get hold of a kernel 2.6.13 without a time machine :-) Should I ignore the error message, or upgrade to the rc version of 2.6.13? 2.6.13-whatever-is-latest. Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
Re: Fedora Inotify Kernels from Red Hat!
On Fri, 2005-07-15 at 12:28 +0200, [EMAIL PROTECTED] wrote: OK, but kernel-desktop iNotify is controlled by /dev/inotify Don't know if it's suitable for beagle then... Can Beagle work with old /dev/inotify device? See my earlier posts. As of yesterday at 1530 EST, Beagle requires the latest system call based inotify. Earlier versions work fine with the device node based inotify. Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
inotify changes landed in CVS.
As of CVS timestamp 2005/07/14 19:28:13, Beagle CVS requires the new syscall-based inotify. Users of SUSE and Fedora can grab the aforementioned kernel packages and ride off into the sunset, content in knowing your distribution cares about you. Folks who roll their own kernel, grab 2.6.13-rc3 or later. Inotify is included, free of charge. Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
inotify system call support.
So, inotify is now in 2.6.13-rc3. The interface changed slightly, to use system calls. The attached patch ports Beagle to the new inotify interface. Robert Love Util/Inotify.cs | 22 -- glue/inotify-glue.c | 58 +++- glue/inotify-syscalls.h | 29 glue/inotify.h | 16 - 4 files changed, 62 insertions(+), 63 deletions(-) Index: Util/Inotify.cs === RCS file: /cvs/gnome/beagle/Util/Inotify.cs,v retrieving revision 1.47 diff -u -u -r1.47 Inotify.cs --- Util/Inotify.cs 6 Jul 2005 22:54:08 - 1.47 +++ Util/Inotify.cs 13 Jul 2005 18:59:58 - @@ -94,7 +94,7 @@ } [DllImport (libinotifyglue)] - static extern void inotify_glue_init (); + static extern int inotify_glue_init (); [DllImport (libinotifyglue)] static extern int inotify_glue_watch (int fd, string filename, EventType mask); @@ -148,7 +148,7 @@ / - static private int dev_inotify = -1; + static private int inotify_fd = -1; static private ArrayList event_queue = new ArrayList (); static Inotify () @@ -163,15 +163,13 @@ if (Environment.GetEnvironmentVariable (BEAGLE_INOTIFY_VERBOSE) != null) Inotify.Verbose = true; - inotify_glue_init (); - - dev_inotify = Syscall.open (/dev/inotify, OpenFlags.O_RDONLY); - if (dev_inotify == -1) -Logger.Log.Warn (Could not open /dev/inotify); + inotify_fd = inotify_glue_init (); + if (inotify_fd == -1) +Logger.Log.Warn (Could not initialize inotify); } static public bool Enabled { - get { return dev_inotify = 0; } + get { return inotify_fd = 0; } } / @@ -350,7 +348,7 @@ // in the wd value changing. // (no need to worry about watched_by_wd being polluted with stale watches) - wd = inotify_glue_watch (dev_inotify, path, mask | base_mask); + wd = inotify_glue_watch (inotify_fd, path, mask | base_mask); if (wd 0) { string msg = String.Format (Attempt to watch {0} failed!, path); throw new IOException (msg); @@ -402,7 +400,7 @@ if (watched.Subscribers.Count 0) return; - int retval = inotify_glue_ignore (dev_inotify, watched.Wd); + int retval = inotify_glue_ignore (inotify_fd, watched.Wd); if (retval 0) { string msg = String.Format (Attempt to ignore {0} failed!, watched.Path); throw new IOException (msg); @@ -423,7 +421,7 @@ watched.Mask |= new_mask; int new_wd; -new_wd = inotify_glue_watch (dev_inotify, +new_wd = inotify_glue_watch (inotify_fd, watched.Path, watched.Mask | base_mask); @@ -489,7 +487,7 @@ int nr; // Will block while waiting for events, but with a 1s timeout. -inotify_snarf_events (dev_inotify, +inotify_snarf_events (inotify_fd, 1, out nr, out buffer); Index: glue/inotify-glue.c === RCS file: /cvs/gnome/beagle/glue/inotify-glue.c,v retrieving revision 1.21 diff -u -u -r1.21 inotify-glue.c --- glue/inotify-glue.c 22 Jun 2005 22:57:11 - 1.21 +++ glue/inotify-glue.c 13 Jul 2005 18:59:58 - @@ -33,17 +33,19 @@ #include errno.h #include sys/ioctl.h #include sys/select.h +#include sys/types.h #include inotify.h +#include inotify-syscalls.h -#define SYSFS_PREFIX /sys/class/misc/inotify +#define PROCFS_PREFIX /proc/sys/fs/inotify -#define SYSFS_MAX_USER_DEVICES SYSFS_PREFIX /max_user_devices -#define SYSFS_MAX_USER_WATCHES SYSFS_PREFIX /max_user_watches -#define SYSFS_MAX_QUEUED_EVENTS SYSFS_PREFIX /max_queued_events +#define PROCFS_MAX_USER_DEVICES PROCFS_PREFIX /max_user_instances +#define PROCFS_MAX_USER_WATCHES PROCFS_PREFIX /max_user_watches +#define PROCFS_MAX_QUEUED_EVENTS PROCFS_PREFIX /max_queued_events /* Inotify sysfs knobs, initialized to their pre-sysfs defaults */ -static int max_user_devices = 8; +static int max_user_instances = 8; static int max_user_watches = 8192; static unsigned int max_queued_events = 256; @@ -67,53 +69,39 @@ } -void +int inotify_glue_init (void) { - static int initialized = 0; - if (initialized) + static int fd = 0; + if (fd) return; - initialized = 1; + fd = inotify_init (); + + read_int (PROCFS_MAX_USER_DEVICES, max_user_instances); + read_int (PROCFS_MAX_USER_WATCHES, max_user_watches); + read_int (PROCFS_MAX_QUEUED_EVENTS, max_queued_events); - read_int (SYSFS_MAX_USER_DEVICES, max_user_devices); - read_int (SYSFS_MAX_USER_WATCHES, max_user_watches); - read_int (SYSFS_MAX_QUEUED_EVENTS, max_queued_events); + return fd; } int inotify_glue_watch (int fd, const char *filename, __u32 mask) { - struct
Re: inotify system call support.
On Wed, 2005-07-13 at 15:52 -0400, Robert Love wrote: So, inotify is now in 2.6.13-rc3. The interface changed slightly, to use system calls. The attached patch ports Beagle to the new inotify interface. Joey had the idea of printing a message if the system call is not defined, thus allowing us to trivially detect a non-inotify kernel and warning the user on how to rectify. Attached patch is updated to do just that. Robert Love Index: Util/Inotify.cs === RCS file: /cvs/gnome/beagle/Util/Inotify.cs,v retrieving revision 1.47 diff -u -u -r1.47 Inotify.cs --- Util/Inotify.cs 6 Jul 2005 22:54:08 - 1.47 +++ Util/Inotify.cs 13 Jul 2005 20:14:23 - @@ -94,7 +94,7 @@ } [DllImport (libinotifyglue)] - static extern void inotify_glue_init (); + static extern int inotify_glue_init (); [DllImport (libinotifyglue)] static extern int inotify_glue_watch (int fd, string filename, EventType mask); @@ -148,7 +148,7 @@ / - static private int dev_inotify = -1; + static private int inotify_fd = -1; static private ArrayList event_queue = new ArrayList (); static Inotify () @@ -163,15 +163,13 @@ if (Environment.GetEnvironmentVariable (BEAGLE_INOTIFY_VERBOSE) != null) Inotify.Verbose = true; - inotify_glue_init (); - - dev_inotify = Syscall.open (/dev/inotify, OpenFlags.O_RDONLY); - if (dev_inotify == -1) -Logger.Log.Warn (Could not open /dev/inotify); + inotify_fd = inotify_glue_init (); + if (inotify_fd == -1) +Logger.Log.Warn (Could not initialize inotify); } static public bool Enabled { - get { return dev_inotify = 0; } + get { return inotify_fd = 0; } } / @@ -350,7 +348,7 @@ // in the wd value changing. // (no need to worry about watched_by_wd being polluted with stale watches) - wd = inotify_glue_watch (dev_inotify, path, mask | base_mask); + wd = inotify_glue_watch (inotify_fd, path, mask | base_mask); if (wd 0) { string msg = String.Format (Attempt to watch {0} failed!, path); throw new IOException (msg); @@ -402,7 +400,7 @@ if (watched.Subscribers.Count 0) return; - int retval = inotify_glue_ignore (dev_inotify, watched.Wd); + int retval = inotify_glue_ignore (inotify_fd, watched.Wd); if (retval 0) { string msg = String.Format (Attempt to ignore {0} failed!, watched.Path); throw new IOException (msg); @@ -423,7 +421,7 @@ watched.Mask |= new_mask; int new_wd; -new_wd = inotify_glue_watch (dev_inotify, +new_wd = inotify_glue_watch (inotify_fd, watched.Path, watched.Mask | base_mask); @@ -489,7 +487,7 @@ int nr; // Will block while waiting for events, but with a 1s timeout. -inotify_snarf_events (dev_inotify, +inotify_snarf_events (inotify_fd, 1, out nr, out buffer); Index: glue/inotify-glue.c === RCS file: /cvs/gnome/beagle/glue/inotify-glue.c,v retrieving revision 1.21 diff -u -u -r1.21 inotify-glue.c --- glue/inotify-glue.c 22 Jun 2005 22:57:11 - 1.21 +++ glue/inotify-glue.c 13 Jul 2005 20:14:23 - @@ -33,17 +33,19 @@ #include errno.h #include sys/ioctl.h #include sys/select.h +#include sys/types.h #include inotify.h +#include inotify-syscalls.h -#define SYSFS_PREFIX /sys/class/misc/inotify +#define PROCFS_PREFIX /proc/sys/fs/inotify -#define SYSFS_MAX_USER_DEVICES SYSFS_PREFIX /max_user_devices -#define SYSFS_MAX_USER_WATCHES SYSFS_PREFIX /max_user_watches -#define SYSFS_MAX_QUEUED_EVENTS SYSFS_PREFIX /max_queued_events +#define PROCFS_MAX_USER_DEVICES PROCFS_PREFIX /max_user_instances +#define PROCFS_MAX_USER_WATCHES PROCFS_PREFIX /max_user_watches +#define PROCFS_MAX_QUEUED_EVENTS PROCFS_PREFIX /max_queued_events /* Inotify sysfs knobs, initialized to their pre-sysfs defaults */ -static int max_user_devices = 8; +static int max_user_instances = 8; static int max_user_watches = 8192; static unsigned int max_queued_events = 256; @@ -67,53 +69,50 @@ } -void +int inotify_glue_init (void) { - static int initialized = 0; - if (initialized) + static int fd = 0; + if (fd) return; - initialized = 1; + fd = inotify_init (); + if (fd 0) { + int _errno = errno; + perror (inotify_init); + switch (_errno) { + case ENOSYS: + fprintf(stderr, Inotify not supported! You need a +2.6.13 kernel or later with CONFIG_INOTIFY +enabled.); + break; + } + } - read_int (SYSFS_MAX_USER_DEVICES, max_user_devices); - read_int (SYSFS_MAX_USER_WATCHES, max_user_watches); - read_int (SYSFS_MAX_QUEUED_EVENTS, max_queued_events); + read_int (PROCFS_MAX_USER_DEVICES, max_user_instances
Re: cp metadata error fix.
On Tue, 2005-05-24 at 10:39 -0400, JP Rosevear wrote: Robert, lets also file this in bnc with the patch (send me the bug number). Yes, sir. Bug #85646. Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
cp metadata error fix.
On Fri, 2005-05-20 at 01:53 -0400, Nat Friedman wrote: It's a pretty scary set of warnings. It would be much nicer if cp said Warning: /media/USB_DISK does not support extended attributes; metadata will be lost. instead of printing one warning per EA. We should follow up with the fileutils people. s/fileutils/coreutils, these days. Anyhow, I looked into it. The error is actually generated in the libattr function attr_copy_file(). I fixed it. We now have $ cp x /var/tmp cp: Metadata will be lost, setting attributes for `/var/tmp/x': Operation not supported Instead of $ cp x /var/tmp cp: setting attribute `user.Beagle.Uid' for `/var/tmp/x': Operation not supported cp: setting attribute `user.Beagle.MTime' for `/var/tmp/x': Operation not supported cp: setting attribute `user.Beagle.IndexTime' for `/var/tmp/x': Operation not supported cp: setting attribute `user.Beagle.Fingerprint' for `/var/tmp/x': Operation not supported Find a patch to libattr attached. I also built updated packages for SUSE 9.3, because I love you all: http://primates.ximian.com/~rml/misc/attr/ Best, Robert Love libattr/attr_copy_file.c |6 +++--- 1 files changed, 3 insertions(+), 3 deletions(-) --- attr-2.4.22/libattr/attr_copy_file.c 2004-11-30 08:34:47.0 -0500 +++ attr/libattr/attr_copy_file.c 2005-05-20 11:49:25.0 -0400 @@ -138,9 +138,9 @@ if (lsetxattr (dst_path, name, value, size, 0) != 0) { const char *qpath = quote (ctx, dst_path); - if (errno == ENOSYS) { -error (ctx, _(setting attributes for %s), - qpath); + if (errno == ENOSYS || errno == EOPNOTSUPP) { +error (ctx, _(setting attributes for %s; + metadata will be lost), qpath); ret = -1; break; /* no hope of getting any further */ } else { ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
updated inotify kernels.
SUSE 9.3 737.inotify.1 kernels: http://primates.ximian.com/~rml/kernel-rml/suse-93-i586/ based on patch 0.23-8: http://www.kernel.org/pub/linux/kernel/people/rml/inotify/v2.6/0.23/ These do NOT break API/ABI: No breakage since the previous. This kernels work with CVS from Mon May 9 17:53:35 EDT 2005 onward and will be the Official Kernel of the forthcoming Beagle 0.0.10 release. Changes are misc. cleanups and fixes and optimizations. Diligently yours, Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
Inotify ABI Breakage.
This is hopefully the last. Latest inotify breaks ABI and API. Beagle as of Mon May 9 17:53:35 EDT 2005 supports the new inotify and is, thus, incompatible. You want patch version 0.23-6 or later: http://www.kernel.org/pub/linux/kernel/people/rml/inotify/v2.6/0.23/ SUSE 9.3 users can grab a kernel package, version 737.inotify.0 or later: http://www.kernel.org/pub/linux/kernel/people/rml/inotify/v2.6/0.23/ The main change to the API is a new IN_ISDIR bit set on all events that pertain to a directory. This allows user-space handy access to the information and allows us to remove the events that had a file/directory delineation, such as IN_CREATE_FILE. Enjoy. Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
Re: updated inotify.
On Wed, 2005-04-27 at 14:23 -0400, Robert Love wrote: Ahem. Also: Updated inotify kernels! SUSE 9.3 packages, version 727.inotify.0: http://primates.ximian.com/~rml/kernel-rml/suse-93-i586/ These use the new inotify 0.23: http://www.kernel.org/pub/linux/kernel/people/rml/inotify/v2.6/0.23/ Now, SUSE 9,3 kernel packages version 727.inotify.1 and kernel patches version 0.23-4 are out. Since the last posting: - Changed the default knob values [*] - Backward compatible IN_ALL_EVENTS flag - Misc. cleanup and optimizations [*] The default number of events is now 8192, from only 512. This should greatly decrease the chance of queue overflow. The default number of devices is also increased, to 128 from 64. This should not affect Beagle, though. The default number of watches, however, is now only 8192, down from 16384. This is because 8192 is a saner maximum for the mainline kernel, in which we are preparing for submission. Beagle users may want to bump this number. The only downside to larger numbers is memory consumption. As always, I recommended setting your own defaults via the exposed sysfs knobs. An init script makes sense for this. One for SUSE systems is attached (it is also checked into CVS at tools/boot.inotify.init). This init script sets your (watches,devices,events) limits to (16k,128,16k). To activate the boot script on SUSE, save it to /etc/init.d/boot.inotify and run as root /sbin/chkconfig --add boot.inotify Best, Robert Love boot.inotify Description: application/shellscript ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
Re: beagled -- ioctl: Bad address
On Fri, 2005-04-15 at 16:13 -0500, Thor Farrow wrote: I've recently installed Beagle on my laptop running Ubuntu Hoary pre-release, and it was running fine until this past week. I'm assuming that one of the Ubuntu package upgrades is causing this, but I am now getting the following error in terminal when attempting to start the beagled: You need to use an earlier Beagle or a newer inotify. Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
updatederest ever inotify kernels.
These kernels are dedicated to Mitch Hedberg, fellow South Floridian and lover of the laugh. Requiescat in Pace. These kernels are ONLY for SUSE 9.3 users using Beagle from CVS. They correspond to the soon-to-be-released 0.22 release of inotify. There are a handful of cleanups and minor fixes in this release, including two important changes: - I converted our internal locking scheme to semaphores. This lets us block more readily and thus clean up a bit of code that was previously contorted. More importantly, no more GFP_ATOMIC allocations. We can sleep whenever we want with whomever we want. Just like my ex-wife. - J.T. spotted a bug in the move code. The internal coalescing code was dropping moves that were duplicate in all values except cookie. We now verify that the cookies match before coalescing. UP kernels: http://primates.ximian.com/~rml/kernel-rml/suse-93-i586/kernel-default-2.6.11.4-717.inotify.4.i586.rpm http://primates.ximian.com/~rml/kernel-rml/suse-93-i586/kernel-default-nongpl-2.6.11.4-717.inotify.4.i586.rpm SMP kernels: http://primates.ximian.com/~rml/kernel-rml/suse-93-i586/kernel-smp-2.6.11.4-717.inotify.4.i586.rpm http://primates.ximian.com/~rml/kernel-rml/suse-93-i586/kernel-smp-nongpl-2.6.11.4-717.inotify.4.i586.rpm Source: http://primates.ximian.com/~rml/kernel-rml/suse-93-i586/kernel-source-2.6.11.4-717.inotify.4.i586.rpm http://primates.ximian.com/~rml/kernel-rml/suse-93-i586/kernel-source-debuginfo-2.6.11.4-717.inotify.4.i586.rpm Best, Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
Re: beagle inotify ABI change
On Sat, 2005-04-02 at 13:19 -0500, Miguel de Icaza wrote: Hi, Miggy. Am using NLD. When I try to install it, I get: linux:~/kernels # rpm -i * error: Failed dependencies: mkinitrd = 1.2 is needed by kernel-default-2.6.11.4-717.inotify.3 Where can I get this package? These kernels are for SUSE 9.3. I would not try running them on NLD, even if you find the missing mkinitrd package. Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
Re: Beagle 0.0.8
On Wed, 2005-03-23 at 20:15 +, James Ogley wrote: Is SuSE 9.3's kernel inotify 0.20-enabled as far as you know? It is not. Packages are available http://primates.ximian.com/~rml/kernel-rml/suse-93-i586/old/ Best, Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
beagle inotify ABI change
Everyone, Starting right now, with current CVS and henceforth all Beagle releases after 0.0.8, Beagle is updated to the inotify 0.21 patch, which introduces an ABI and API change. Hear ye, hear ye: *** Beagle 0.0.8 requires inotify 0.20 *** Beagle CVS (0.0.9 and on) requires inotify 0.21 Kernels for SUSE 9.3 are available: http://primates.ximian.com/~rml/kernel-rml/suse-93-i586/ Kernels for SUSE 9.1, SUSE 9.2, SLES9, and NLD9 are available: http://primates.ximian.com/~rml/kernel-rml/suse-92-i586/ Kernel patches: http://www.kernel.org/pub/linux/kernel/people/rml/inotify/v2.6/0.21/ If you use CVS, please make sure you use these updated packages/patches. If you use the 0.0.8 release or what is included in SUSE 9.3, please use the older kernels or 0.20 patch. Thanks. Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
Re: Beagle 0.0.8
On Wed, 2005-03-23 at 22:00 +, James Ogley wrote: It is not. Packages are available Presumably I'll need to rebuild beagle too? No, not for the kernels to which I linked. For the newer kernels, yes. But the kernels in old/ are for the Beagle we shipped in 9.3 Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
autofs, anyone?
My minions, Do any of you use autofs? If so, I need you! Andrew Morton is interested in what happens if an inotify watch triggers an automount. Finding out should be easy. Say you have an automount at /foo/bar. Then (in Util in the beagle src) ./inotify-test -r /foo/bar and let me know what happens. If the automount is in your /home, and Beagle works, let me know. Possible actions are (a) Nothing (b) The automount is triggered and mounted. (c) Some sort of failure. I hope for (b). Thanks, Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
Re: totally new inotify.
On Fri, 2005-03-04 at 14:17 -0500, Robert Love wrote: For lucky SUSE users, Grand Unified Kernels version 747.inotify.1 for SUSE 9.1, SUSE 9.2, SLES9, and NLD9 are available at http://primates.ximian.com/~rml/kernel-rml/unified/ For users of other distributions, or the rough riders amongst us who compile their own destiny, patches are available at http://www.kernel.org/pub/linux/kernel/people/rml/inotify/v2.6/0.20 The kernel packages are updated to 747.inotify.2 and the patch is updated to 0.20-2. Fixes, cleanups, and optimizations. Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
Re: Giant memory leaks still there
On Thu, 2005-02-24 at 15:37 -0500, Joe Shaw wrote: --debug-memory will tell you the size of the daemon process. It uses VmSize but I'm about to commit something that will display VmRss instead. Is not VmSize better for seeing leaks? Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
updatedest inotify kernels
On Tue, 2005-02-08 at 14:59 -0500, Robert Love wrote: I have updated inotify kernels available. THESE DO NOT WORK WITH VERSIONS OF BEAGLE OTHER THAN RECENT CVS. SPECIFICALLY, THESE DO NOT WORK WITH BEAGLE 0.0.5. 737.inotify.3 is out. The major change is that the out-of-tree drivers (such as Atheros) are again included. I changed my build environment and missed them in the last build. Sorry! 737.inotify.4 is out. I am on a roll. Changes include some performance optimizations and invariants from the locking rewrite unrelated to the locking rewrite. SUSE Grand Unified Kernels (SUSE 9.1, 9.2, NLD9, and SLES9) are here http://primates.ximian.com/~rml/kernel-rml/unified/ The patch is here http://www.kernel.org/pub/linux/kernel/people/rml/inotify/v2.6/0.18/inotify-0.18-rml-2.6.10-16.patch The definition of irony is here http://www.google.com/search?hl=enlr=q=define%3AironybtnG=Search Note: The first definition is _not_ a definition of irony. All my love, Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers
required inotify kernel upgrade
Everyone, please read this, if you can read. As of right now, Beagle-CVS is incompatible with the previous inotify kernels and patches. Inotify has moved to a dynamically allocated and sized filename payload. This means that the average inotify event is significantly smaller, yet inotify can support filenames of any size. The introduction of dynamic sized filenames required a bit of effort, due to alignment issues requiring null padding, but the patch has been out there for awhile (inotify 0.18) and I just finished the Beagle work. SUSE 9.1, 9.2, NLD9, and SLES9 users can use the following Grand Unified Kernels: http://primates.ximian.com/~rml/kernel-rml/suse-92-i586/ Roughriders who compile from scratch can use the following patch, against 2.6.10: http://www.kernel.org/pub/linux/kernel/people/rml/inotify/v2.6/0.18/inotify-0.18-rml-2.6.10-9.patch Hopefully the other distros who build kernel packages will update to 0.18 soon. The patch has been out for awhile now. This is, we hope, the last (major) ABI/API change to Inotify. Happy hacking, Robert Love ___ Dashboard-hackers mailing list Dashboard-hackers@gnome.org http://mail.gnome.org/mailman/listinfo/dashboard-hackers