Bug 1201749

Description of problem:

For some reason the 'coresize' field in the 'tasks' table is always NULL, which must be a bug.

$ sqlite3 /cores/retrace/tasks/stats.db
sqlite> .dump tasks
...

INSERT INTO "tasks" VALUES(7350,854078172,'kernel','2.6.32-220.17.1.el6','x86_64',1426209353,486,NULL,6);
INSERT INTO "tasks" VALUES(7351,480791831,'kernel','2.6.32-220.17.1.el6','x86_64',1426209424,468,NULL,6);
INSERT INTO "tasks" VALUES(7352,937645551,NULL,NULL,NULL,1426213076,601,NULL,7);
INSERT INTO "tasks" VALUES(7353,286770388,'kernel','2.6.18-371.12.1.el5','x86_64',1426227219,77,NULL,6);
INSERT INTO "tasks" VALUES(7354,939050538,'kernel','2.6.32-358.el6','x86_64',1426231351,2026,NULL,6);
INSERT INTO "tasks" VALUES(7355,791138625,'kernel','2.6.32-358.el6','x86_64',1426231396,2420,NULL,6);
INSERT INTO "tasks" VALUES(7356,951548981,'kernel','2.6.32-358.el6','x86_64',1426231428,3060,NULL,6);
INSERT INTO "tasks" VALUES(7357,174468018,'kernel','2.6.32-504.3.3.el6','x86_64',1426235617,110,NULL,6);
INSERT INTO "tasks" VALUES(7358,527768478,'kernel','2.6.32-431.20.3.el6','x86_64',1426247554,33,NULL,6);
CREATE TABLE success(taskid REFERENCES tasks(id), pre NOT NULL, post NOT NULL,
              rootsize NOT NULL);
CREATE TABLE packages(id INTEGER PRIMARY KEY AUTOINCREMENT,
               name NOT NULL, version NOT NULL);
CREATE TABLE packages_tasks(pkgid REFERENCES packages(id),
                     taskid REFERENCES tasks(id));
CREATE TABLE buildids(taskid REFERENCES tasks(id), soname, buildid NOT NULL);
CREATE TABLE reportfull(requesttime NOT NULL, ip NOT NULL);
DELETE FROM sqlite_sequence;
INSERT INTO "sqlite_sequence" VALUES('tasks',7358);
COMMIT;
sqlite> .schema tasks
CREATE TABLE tasks(id INTEGER PRIMARY KEY AUTOINCREMENT, taskid, package, version,
      arch, starttime NOT NULL, duration NOT NULL, coresize, status NOT NULL);



Version-Release number of selected component (if applicable):
retrace-server-1.12-3.el6.noarch


How reproducible:
Any retrace-server task has this problem.


Steps to Reproduce:
Dump out any record in the tasks table and look at 'coresize'

Actual results:
NULL coresize

Expected results:
Non-null coresize


Additional info:
I briefly looked at this but the error wasn't obvious.  Probably turn on some debugging and run it and the problem may be obvious.  At first I thought maybe the filename was just wrong - I didn't know the vmcore was named 'coredump' at one point.  However, this looks ok since we use it later to set the 'arch' and this is always set properly in the database, or so it seems.

See /usr/bin/retrace-server-worker and start_retrace() 

def start_retrace():
    crashdir = os.path.join(task.get_savedir(), "crash")
    corepath = os.path.join(crashdir, "coredump")  <------------------ is that right?

    try:
        stats["coresize"] = os.path.getsize(corepath)
    except:
        pass

   if cmdline.arch:
        log_debug("Using architecture from command line: %s" % cmdline.arch)
        arch = cmdline.arch
    else:
        # read architecture from coredump
        arch = guess_arch(corepath)

        if not arch:
            log_error("Unable to determine architecture from coredump")
            fail()

        log_debug("Determined architecture: %s" % arch)

    stats["arch"] = arch