Hello Richard, On Mon, Feb 07, 2011 at 07:02:26PM -0500, Richard Hipp wrote: > 2011/2/7 Lluís Batlle i Rossell <virik...@gmail.com> > > > On Mon, Feb 07, 2011 at 05:55:08PM -0500, Richard Hipp wrote: > > > I haven't done an official build of Fossil on a long while. So I'll try > > to > > > do one on Wednesday. Please let me know if you see any serious problems > > > with the current tip of trunk. > > > > > > Note that I am unable to reproduce the sync problem reported in > > > http://www.fossil-scm.org/fossil/info/fbea61caf0 despite considerable > > > effort. > > > > Hello, > > > > it's me reporting this problem. We experience this in all our machines, > > which use fossil versions all from this year. > > > > We did not see any of this until we enabled the delta manifests, around > > december. We do not commit baseline manifests at will; it is fossil > > thinking it > > is time to do so. > > > > We use mostly linux x86_64, and also some cygwin fossil windows builds, at > > a > > commit rate of around 10 commits per day, with quite enough files: > > $ fossil ls | wc -l > > 2930 > > > > All coworkers experenienced this, all synchronising against a fossil served > > by > > cgi. The fossil server is not that new; maybe it dates from late november. > > Does > > it matter? > > > > Yes. Older Fossils don't know how to deal with delta manifests. (On the > other hand, delta manifests should never be created unless you explicitly > ask for them at least once. Once at least one delta-manifest is in the > respository, Fossil will feel free to create more, but it will never create > the first one without being asked.)
I know. We introduced the first delta manifest on December, and we work with deltas since then in all workstations (around 10 workstations, 5 users). > > So if you have asked for a delta-manifest, please update all Fossil binaries > to the latest. Please let me know if this does not fix the problem. All fossil versions we use are aware of delta manifests, even that running as CGI. Simply the CGI is older than the versions we use in the workstations. In fact, the CGI has *never* shown any problem like that of not going beyond new baseline manifests. So it goes like this: Workstation A: - Adds some delta manifests, autosynching with the CGI - At some points, adds a baseline manifest (B1, let's name), autosynching. - Adds more delta manifests, autosynching. Server (CGI, running fossil 2011-01-01, aware of deltas): - It always shows proper branches in the timeline, although there we don't have any opened repository. Workstation B or any other (latest fossil version): - Autosynhing from the CGI, it interchanges all artifacts. - Running "fossil update trunk", or looking at the timeline, the last baseline checkin (B1) is the last shown, or the last updated to. Committing anything and autosynching, creates a new leaf on trunk (seen in the CGI), although it thinks it has only one leaf. - "fossil rebuild" puts the timeline shown as the server, brings in the double leaf, and all delta manifests that were 'hidden' until that moment. If it is Workstation B commiting the baseline manifest, then all others (A, C, ...) experience the same problem, and we need a 'rebuild in them'. The server always shows properly the last state with all deltas taken into account. This happens with every baseline checkin. We noticed the first baseline checkin (after introducing deltas) at the beginning of January. I forgot to mention this CGI server working properly, in previous letters. And I said it was from late november; bad memory. The server is from 2011-01-01, i686-linux. I understand this is hard to track... I would send you a '.fossil' file of a repository in 'that state', but we are managing closed source. Can you suggest at what sqlite tables I should look at? I don't have all the workstations at the 'last checkin' right now, so I imagine I can find one with "needs a rebuild". I agree the error looks hard to find. Thank you, Lluís. > > It looks like all fossil 'clients' get always the changes on sync, but they > > are > > not considered in the checkin DAG. Only after 'rebuild' they become part of > > the > > DAG. > > > > Tomorrow I can try to reproduce this in a new repository. > > > > Who is here using delta manifests in their repositories? Anyone else? I > > know > > fossil-scm.org does not use delta manifests. > > > > Regards, > > Lluís. > > _______________________________________________ > > fossil-users mailing list > > fossil-users@lists.fossil-scm.org > > http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users > > > > > > -- > D. Richard Hipp > d...@sqlite.org > _______________________________________________ > fossil-users mailing list > fossil-users@lists.fossil-scm.org > http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users _______________________________________________ fossil-users mailing list fossil-users@lists.fossil-scm.org http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users
