[fossil-users] commits from host A sometimes not seen on B

[Michai Ramakers]

[ following up on oldish thread from mid August this year, with the
same subject ]

I've just seen this issue here, albeit in slightly another form
perhaps (2 systems 'S' (server, amd64 netbsd) and 'C' (client, WinXP
pro). 'S' makes a repo available using CGI/inetd, and I work on 'S'
locally as well, with a local repo cloned from
"http://S/myrepo.fossil". Autosync is on for local repos on 'C' and
'S'.

I did the following:

1) check-in on C
2) update local repo on S
3) local repo on S doesn't show recent check-in, as per 'fossil timeline'
4) CGI-shared repo on S does show the check-in as per web interface timeline
5) make copies of C local repo, S local repo, S CGI-shared repo for analysis
6) fetch/build latest fossil, build, and run 'fossil sync --verily' on
S local repo
7) all is fine; recent check-in is seen

Stupidly, I didn't try a plain "fossil sync" or "update" after
building. I can still do this by reverting to the saved archives,
tonight (probably).

I can make the 3 repos available for debugging (for Richard) on
request, but I have a pretty thin uplink (but lots of space).

Michai

[Stephan Beal]

Post by Michai Ramakers
2) update local repo on S
3) local repo on S doesn't show recent check-in, as per 'fossil timeline'
4) CGI-shared repo on S does show the check-in as per web interface timeline

What is the time span between 2==>3 and 3==>4? i ask because i have one
particular system which often takes several seconds to flush changes to
disk. e.g. after compiling a new fossil binary, i sometimes have to wait 3
seconds before running the binary actually gets the new copy. i'm
_speculating_ that if the timespan is short, then maybe it has to do with
drive caching. Otherwise i'm just as confused as everyone else is regarding
this behaviour.

[Michai Ramakers]

Post by Stephan Beal

Post by Michai Ramakers
2) update local repo on S
3) local repo on S doesn't show recent check-in, as per 'fossil timeline'
4) CGI-shared repo on S does show the check-in as per web interface timeline

What is the time span between 2==>3 and 3==>4? i ask because i have one
particular system which often takes several seconds to flush changes to
disk. e.g. after compiling a new fossil binary, i sometimes have to wait 3
seconds before running the binary actually gets the new copy. i'm
_speculating_ that if the timespan is short, then maybe it has to do with
drive caching. Otherwise i'm just as confused as everyone else is regarding
this behaviour.

I don't think that is it; when viewing the timeline on the server, I do see
the check-in. It seems to me that the local repo on 'S' is not syncing from
the CGI-available repo on 'S'.

Michai

[Michai Ramakers]

Post by Michai Ramakers
I don't think that is it; when viewing the timeline on the server, I do see
the check-in.

"when viewing the timeline on the server, using the web interface ..."

Michai

[Michai Ramakers]

Post by Michai Ramakers
2) update local repo on S
3) local repo on S doesn't show recent check-in, as per 'fossil timeline'
4) CGI-shared repo on S does show the check-in as per web interface timeline

What is the time span between 2==>3 and 3==>4? i ask because i have one particular system which often takes several seconds to flush changes to disk. e.g. after compiling a new fossil binary, i sometimes have to wait 3 seconds before running the binary actually gets the new copy. i'm _speculating_ that if the timespan is short, then maybe it has to do with drive caching. Otherwise i'm just as confused as everyone else is regarding this behaviour.

time span between 2/3 >= 10 sec (I did 'fossil timeline' approx 10 sec
after I noticed the update showed the message for an older check-in).

time span between 3/4 also >= 10 sec.

Michai

[Richard Hipp]

Post by Michai Ramakers
I can make the 3 repos available for debugging (for Richard) on
request, but I have a pretty thin uplink (but lots of space).

How big are the repositories? Can you email them?

[Michai Ramakers]

Post by Richard Hipp

Post by Michai Ramakers
I can make the 3 repos available for debugging (for Richard) on
request, but I have a pretty thin uplink (but lots of space).

How big are the repositories? Can you email them?

400 something MB (per repo).

I'd rather put them up for http on my host here, and email you the
URLs. I can do that in an hour, say.

Michai

[Richard Hipp]

Post by Michai Ramakers

Post by Richard Hipp

Post by Michai Ramakers
I can make the 3 repos available for debugging (for Richard) on
request, but I have a pretty thin uplink (but lots of space).

How big are the repositories? Can you email them?

400 something MB (per repo).
I'd rather put them up for http on my host here, and email you the
URLs. I can do that in an hour, say.

OK