in case someone else changed since we read it.
Hopefully this will fix a race condition
where WU results get sent to different HR classes.
(Alternatively we could use transactions,
or acquire the semaphore during read/update,
but this could impact performance).
svn path=/trunk/boinc/; revision=14710
- server: limit global prefs mod time to now
These changes address the situation where a server
sends out prefs with mod time far in the future,
and there's no way to undo them
svn path=/trunk/boinc/; revision=14664
use most recently created host with same domain name,
IP addr, OS name, and CPU model, if one exists.
(from Kevin Reed)
svn path=/trunk/boinc/; revision=14597
It didn't work on a host's first scheduler RPC
(when it passes an auth but no host ID).
How do we look up user in this case?
The weak auth is not stored explicitly in the DB.
Solution: include the user ID in the weak auth.
(this invalidates existing weak auths)
svn path=/trunk/boinc/; revision=14546
This is for debugging apps (currently works only in Unix).
What it does: when running an app,
the client does everything except actually fork/exec the app,
i.e. it sets up the slot dir, creates shared mem segment etc.
It then continues as if the app were actually running,
and you can then manually run your app under a debugger
in the slot directory.
Note: the client won't notice the termination of your app.
- API, Unix: in situations where the timer thread wants to exit
(e.g. it notices a missing heartbeat).
don't directly call boinc_exit(),
since this touches data structures that the worker thread
may be using concurrently.
Instead, set a flag telling the worker thread to call boinc_exit()
(which it will do from its signal handler)
This is an attempt to fix problems reported by Bernd;
I haven't tested it.
- scheduler: add config flag for uploading usage data
- web: show account key and weak account key on user page
- added some code for multithread support (not finished)
api/
boinc_api.C
svn path=/trunk/boinc/; revision=14542
Previously it was called from both timer and worker thread,
with no synchronization. Bad.
boinc_finish(): sleep for 2 seconds, enough to ensure that
the timer thread sends the final messages.
- API: a compile flag GETRUSAGE_IN_TIMER_THREAD selects
how to get CPU time; by default set only for Mac OS X
- scheduler: cap #CPUs scaling for <max_wus_per_host> at 8
svn path=/trunk/boinc/; revision=14495
- web: add app selection to example project-specific prefs
(commented out by default)
- web: in prefs, flag too-low numeric values as errors,
rather than just setting to min value
- web: fix bug when add new venue from "columns" view
svn path=/trunk/boinc/; revision=14442
Each account has an associated weak account key.
The user can see it by visiting weak_auth.php.
If you put the weak account key in the account file on a computer
(instead of the regular account key)
the computer will attach to the account.
But the weak account key cannot be used to log in
to the account on the web, or to do web RPCs, or anything else.
Note: this involves a scheduler change.
svn path=/trunk/boinc/; revision=14367
- user web: don't show team forum in Q&A page
- scheduler: code (commented out) to recognize when
a sched request is from an existing host
(same IP/domain/RAM size) and reuse the host record;
from Kevin Reed
svn path=/trunk/boinc/; revision=14283
1) a WU is marked as ready for assimilation and has no errors;
2) it has no canonical result
In this case, the assimilate handler gets called anyway,
typically with the canonical result of the previous WU as arg.
Note: this situation doesn't arise normally;
it might happen if some results are deleted accidentally.
The fix:
- identify this situation, and set the WU.error_mask to a new code
(WU_ERROR_NO_CANONICAL_RESULT)
- zero out the "canonical_result" variable passed to the handler,
so even if the handler fails to check wu.error_mask,
at least it won't assimilate the same result twice.
Thanks to Hendrik Verhoek for finding this bug.
- DB schema: team table type is MyISAM, not InnoDB
svn path=/trunk/boinc/; revision=13938
and access to worker_thread_ru.
This was being used in the worker signal handler,
which is bad because pthread_mutex_lock() can allocate memory.
I don't think this matters; at worst we might get
CPU times off by < 1 sec.
svn path=/trunk/boinc/; revision=13844
in wu_is_infeasible(), check whether host type is unknown
before seeing if WU is already committed to different type
svn path=/trunk/boinc/; revision=13777
"task" instead of "result";
"computer" instead of "host";
"scheduler request" instead of "scheduler RPC"
"access to Internet" instead of "access to reference web site"
svn path=/trunk/boinc/; revision=13774
that caused it to report "volume full" erroneously
- added --help option to some backend programs;
gives general info and directs to web page
file_upload_handler.C
svn path=/trunk/boinc/; revision=13685
update_versions to add version 6 apps.
It looks for API_VERSION string in main executable,
adds the API version to the app_version XML,
and sets min_core_version to 6 for version 6+ apps
- API: include API_VERSION string
- convert tabs to spaces here and there
- scheduler: parse unused elements in <net_stats>
- ops/show_log.php: if no URL args, just show form (fixes#415)
- client: parse and store api_version (not used yet)
svn path=/trunk/boinc/; revision=13627
of file deletion. If you set file_deletion_strategy to 0 (default)
then this turns off file deletion in the locality scheduler.
Set file_deletion_strategy to 1 to get current behavior. Other
values are not defined. I'll document this in the Wiki.
sched/
sched_config.h
sched_config.C
sched_locality.C
svn path=/trunk/boinc/; revision=13594
in get_file_size() RPC, check if volume has < 1MB available space
and if so return transient error.
This prevents client from trying upload,
which apparently sends entire file even if
handler fails at the beginning
- API, Win: remove use of GetTickCount(); use dtime() instead
- use new BOINC logo in sample project home page
svn path=/trunk/boinc/; revision=13574
tells the client to use this as the data directory
- scheduler: improve the message telling the client that
more disk or memory is needed;
tell them the minimum amount needed to
send any of the jobs rejected,
rather than the amount needed for the first job rejected
- manager: fix text in "connect now" dialog
svn path=/trunk/boinc/; revision=13387
In principle 0660 should work (assuming apache belongs to admin group)
but apparently on some Linux systems it doesn't,
and the feeder gets "can't attach to shared mem" errors.
- scheduler: print uid/gid info if can't attach to shared mem
- user web: fix bug in team founder transfer form
svn path=/trunk/boinc/; revision=13255
David Anderson