Julian Foad wrote on Wed, Dec 08, 2010 at 13:30:42 +:
> I (Julian Foad) wrote:
> [...]
> > Studying the FSFS source code for the issues raised in this thread has
> > given me confidence that it seems to be doing the right thing, in
> > practice, at the moment.
> >
> > In 1043360 I added some c
Julian Foad wrote on Tue, Dec 07, 2010 at 18:13:09 +:
> On Tue, 2010-12-07, Daniel Shahaf wrote:
> > Julian Foad wrote on Tue, Dec 07, 2010 at 15:27:05 +:
> > > (Quoting and replying to two emails at once.)
> > >
> > > First: I'm assuming that no process will ever do packing without gettin
I (Julian Foad) wrote:
[...]
> Studying the FSFS source code for the issues raised in this thread has
> given me confidence that it seems to be doing the right thing, in
> practice, at the moment.
>
> In 1043360 I added some comments about how to remove one source of
> fragility in using the cache
On Wed, 2010-12-08, Daniel Shahaf wrote:
> [ continuing the non-linear replies ]
>
> Daniel Shahaf wrote on Tue, Dec 07, 2010 at 18:51:12 +0200:
> > Julian Foad wrote on Tue, Dec 07, 2010 at 15:27:05 +:
> > > On Tue, 2010-12-07 at 12:13 +0200, Daniel Shahaf wrote:
> > > > Philip Martin wrote o
[ continuing the non-linear replies ]
Daniel Shahaf wrote on Tue, Dec 07, 2010 at 18:51:12 +0200:
> Julian Foad wrote on Tue, Dec 07, 2010 at 15:27:05 +:
> > On Tue, 2010-12-07 at 12:13 +0200, Daniel Shahaf wrote:
> > > Philip Martin wrote on Tue, Dec 07, 2010 at 09:49:40 +:
> > > > Julian
On Tue, 2010-12-07, Daniel Shahaf wrote:
> Julian Foad wrote on Tue, Dec 07, 2010 at 15:27:05 +:
> > (Quoting and replying to two emails at once.)
> >
> > First: I'm assuming that no process will ever do packing without getting
> > the exclusive write lock. Can you confirm that for me? If th
Julian Foad wrote on Tue, Dec 07, 2010 at 15:27:05 +:
> (Quoting and replying to two emails at once.)
>
> First: I'm assuming that no process will ever do packing without getting
> the exclusive write lock. Can you confirm that for me? If that's
> false, all my reasoning would be bogus.
>
(Quoting and replying to two emails at once.)
First: I'm assuming that no process will ever do packing without getting
the exclusive write lock. Can you confirm that for me? If that's
false, all my reasoning would be bogus.
On Tue, 2010-12-07 at 12:13 +0200, Daniel Shahaf wrote:
> Philip Marti
Philip Martin wrote on Tue, Dec 07, 2010 at 09:49:40 +:
> Julian Foad writes:
>
> > On Mon, 2010-12-06 at 18:44 +, Philip Martin wrote:
> >> Julian Foad writes:
> >>
> >> > I'm going to drop this "Remove the re-try logic from
> >> > svn_fs_fs__path_rev_absolute()" follow-up patch, as I
Julian Foad wrote on Tue, Dec 07, 2010 at 09:40:07 +:
> On Mon, 2010-12-06 at 18:44 +, Philip Martin wrote:
> > Julian Foad writes:
> >
> > > I'm going to drop this "Remove the re-try logic from
> > > svn_fs_fs__path_rev_absolute()" follow-up patch, as I don't want to get
> > > into check
Julian Foad writes:
> On Mon, 2010-12-06 at 18:44 +, Philip Martin wrote:
>> Julian Foad writes:
>>
>> > I'm going to drop this "Remove the re-try logic from
>> > svn_fs_fs__path_rev_absolute()" follow-up patch, as I don't want to get
>> > into checking or messing with the txn-correctness o
On Mon, 2010-12-06 at 18:44 +, Philip Martin wrote:
> Julian Foad writes:
>
> > I'm going to drop this "Remove the re-try logic from
> > svn_fs_fs__path_rev_absolute()" follow-up patch, as I don't want to get
> > into checking or messing with the txn-correctness of FSFS now. If
> > Daniel or
Julian Foad writes:
> I'm going to drop this "Remove the re-try logic from
> svn_fs_fs__path_rev_absolute()" follow-up patch, as I don't want to get
> into checking or messing with the txn-correctness of FSFS now. If
> Daniel or anyone else wants to pursue it, I'd be glad to help.
I thought the
id not
> > > > eliminate
> > > > the window of opportunity for the caller to use an out-of-date result.
> > > >
> > > > See the email thread
> > > > <http://svn.haxx.se/dev/archive-2010-12/0019.shtml>,
> > > > subject
Stefan Fuhrmann wrote on Sun, Dec 05, 2010 at 23:45:50 +0100:
> On 01.12.2010 14:16, Daniel Shahaf wrote:
>> Julian Foad wrote on Wed, Dec 01, 2010 at 12:32:45 +:
>>> On Wed, 2010-12-01, stef...@apache.org wrote:
Port (not merge) a fix for a FSFS packing race condition from the
perfor
On 02.12.2010 08:18, Daniel Shahaf wrote:
On Wed, 2010-12-01, stef...@apache.org wrote:
Modified: subversion/trunk/subversion/libsvn_fs_fs/fs_fs.c
URL:
http://svn.apache.org/viewvc/subversion/trunk/subversion/libsvn_fs_fs/fs_fs.c?rev=1040832&r1=1040831&r2=1040832&view=diff
=
On 01.12.2010 14:16, Daniel Shahaf wrote:
Julian Foad wrote on Wed, Dec 01, 2010 at 12:32:45 +:
On Wed, 2010-12-01, stef...@apache.org wrote:
Port (not merge) a fix for a FSFS packing race condition from the
performance branch to /trunk: There is a slight time window
between finding the nam
Julian Foad wrote on Thu, Dec 02, 2010 at 15:59:34 +:
> On Thu, 2010-12-02 at 17:40 +0200, Daniel Shahaf wrote:
> > Julian Foad wrote on Thu, Dec 02, 2010 at 15:34:48 +:
> > > First step: this patch fixes the comments. Good to commit?
> > >
> > > [[[
> > > Index: subversion/libsvn_fs_fs/f
On Thu, 2010-12-02 at 17:40 +0200, Daniel Shahaf wrote:
> Julian Foad wrote on Thu, Dec 02, 2010 at 15:34:48 +:
> > First step: this patch fixes the comments. Good to commit?
> >
> > [[[
> > Index: subversion/libsvn_fs_fs/fs_fs.c
> > ===
Julian Foad wrote on Thu, Dec 02, 2010 at 15:34:48 +:
> First step: this patch fixes the comments. Good to commit?
>
> [[[
> Index: subversion/libsvn_fs_fs/fs_fs.c
> ===
> --- subversion/libsvn_fs_fs/fs_fs.c (revision 1041350)
Daniel Shahaf wrote:
> Julian Foad wrote on Thu, Dec 02, 2010 at 14:33:19 +:
> > I note that the following comment in pack_shard() is not quite right:
> >
> > /* Update the min-unpacked-rev file to reflect our newly packed shard.
> >* (ffd->min_unpacked_rev will be updated by open_pack_o
for the caller to use an out-of-date result.
> > >
> > > See the email thread <http://svn.haxx.se/dev/archive-2010-12/0019.shtml>,
> > > subject "Re: svn commit: r1040832 - Port a fix for a FSFS packing race".
> > >
> > > * subversion
chive-2010-12/0019.shtml>,
> > subject "Re: svn commit: r1040832 - Port a fix for a FSFS packing race".
> >
> > * subversion/libsvn_fs_fs/fs_fs.c
> > (svn_fs_fs__path_rev_absolute): Remove the re-try logic.
> >
> > * subversion/libsvn_fs_fs/fs_f
patch for this, please?
>
> - Julian
>
>
> Remove the re-try logic from svn_fs_fs__path_rev_absolute(). Since
> r1040832, all its callers correctly account for the possibility of an
> out-of-date result due to a concurrent packing operation.
>
> The re-try logic was
e now (since all
> callers either run under the write lock or retry)...
Can you check the attached patch for this, please?
- Julian
Remove the re-try logic from svn_fs_fs__path_rev_absolute(). Since
r1040832, all its callers correctly account for the possibility of an
out-of-date result due to a
> On Wed, 2010-12-01, stef...@apache.org wrote:
> > Modified: subversion/trunk/subversion/libsvn_fs_fs/fs_fs.c
> > URL:
> > http://svn.apache.org/viewvc/subversion/trunk/subversion/libsvn_fs_fs/fs_fs.c?rev=1040832&r1=1040831&r2=1040832&view=diff
> >
Julian Foad wrote on Wed, Dec 01, 2010 at 12:32:45 +:
> On Wed, 2010-12-01, stef...@apache.org wrote:
> > Port (not merge) a fix for a FSFS packing race condition from the
> > performance branch to /trunk: There is a slight time window
> > between finding the name of a rev file and actually ope
On Wed, 2010-12-01, stef...@apache.org wrote:
> Port (not merge) a fix for a FSFS packing race condition from the
> performance branch to /trunk: There is a slight time window
> between finding the name of a rev file and actually open it. If
> the revision in question gets packed just within this w
28 matches
Mail list logo
