Thanks; I committed something equivalent to this.
I'm confused about how the wrapper *ever* worked on Unix;
wouldn't it treat every sub-task as a failure?
-- David
On 23-Jul-2015 12:26 PM, Sébastien Lauzier wrote:
Hi,
I'm currently working on a project that's using the wrapper on linux. I made
some changes to the wrapper specifically for the app it's calling. After those
changes, when I started testing the wrapper was finishing in error and logging
an error code for the app it called. The strange thing was that the app started
fine, the log files showed that everything worked correctly, no warning or
error. I ended up reading the doc for waitpid() used in the poll method of the
TASK class and found that WEXITSTATUS(status) return the exit status of the
child. The value returned from WEXITSTATUS(status) to determine if the child
returned an error or finish successfully and it matched expected value for the
different test I ran.
I didn't do extensive testing, only on 2 different computer running Ubuntu
14.04 64b.
I'm surprise nobody encountered that prior to us and I wonders if it's the way
we are using the wrapper that causing this.
I made the following changes in wrapper.cpp, feedback, review would be
appreciated.
if (task.poll(status)) {
+ //get the child exit status
+ if (WIFEXITED(status)) {
+ status = WEXITSTATUS(status)
+ }
if (status) {
....
I can provide a proper patch if needed.
Seb
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.
