shared write activity crash on joined machine when changing current active activity

[erikos]

Reproduce:

A: shares write B: joins and switches to another activity (changes current active activity)

Tested on 687 and 682 with gabble and salut.

Logs are here: http://olpc.scheffers.net/olpc/browse.tcl?p=20080123-215041-85.178.199.253

There is nothing in the write logs but i collected the attached one from another try when debug was not enabled.

[erikos]

write log member change

[erikos]

Oups the attached one org.laptop.AbiWordActivity-1.log was not the right one. But i hope it can reproduce in the way i described.

[erikos]

The log gives a hint that setting the active activity does fail. The error handler for service.SetActive gets called.

sugar/src/model/homemodel.py:210:        logging.error("set_active() failed: %s" % err)

I could only reproduce this for a shared write activity so far. Reproduced on 691 as well.

[erikos]

Test a) Change active activity The call SetActive(False) for Write times out 'Message did not receive a reply'. Write crashes.

Test b) Keep as plain text The log is attached. Write crashes.

In both cases sharing did work fine before the action.

The test above shows that write will not only crash when changing the current active activity.

[erikos]

Keep as plain text in joined Write

[erikos]

I tested with a few other activities (browse, record, memorize, chat, calculate) and it only happens in write. gabble and salut as well produce the same behavior.

Attaching strace to the process on the machine which joined it looks like Write segfaulted.

[tomeu]

Stack trace:

Program received signal SIGSEGV, Segmentation fault.
0xb4621dc8 in PD_Document::saveAs (this=0x8a48800, output=0x8eadf08, ieft=10, cpy=true, expProps=0x0) at pd_Document.cpp:608
608             return _saveAs(output, ieft, cpy, expProps);
(gdb) bt full
#0  0xb4621dc8 in PD_Document::saveAs (this=0x8a48800, output=0x8eadf08, ieft=10, cpy=true, expProps=0x0)
    at pd_Document.cpp:608
No locals.
#1  0xb43468de in abi_widget_get_content (w=0x8377298, extension_or_mimetype=0x84abe94 ".txt", exp_props=0x0, 
    iLength=0xbfa14200) at abiwidget.cpp:1201
        ieft = 10
        sink = (GsfOutputMemory *) 0x8eadf08
        result = -47600495
        size = 139141824
        ibytes = (const guint8 *) 0x8c0f35c "\001"
        szOut = (gchar *) 0x1a141f4 <Address 0x1a141f4 out of bounds>
#2  0xb48efe4f in _wrap_abi_widget_get_content (self=0x8a87144, args=0xb7d1602c, kwargs=0x8c0f35c) at ./pyabiword.override:54
        bytes = <value optimized out>
        extension_or_mimetype = (gchar *) 0x84abe94 ".txt"
        exp_props = (gchar *) 0x0
        iLength = 146862940
        kwlist = {0xb48f1e0e "extension_or_mimetype", 0xb48f1e24 "exp_props", 0x0}

[erikos]

#5524 stated that this was an issue on 653 already. Make this a blocker since #5524 was a blocker.

[tomeu]

Submitted a patch upstream:

http://bugzilla.abisource.com/show_bug.cgi?id=11421

[cscott]

Patch has been committed upstream. How do we turn the crank to get a new version of the write activity? (Or is there an abiword rpm that needs to be updated?)

[marco]

We need uwog to build rpms.

[uwog]

I'm working on getting 2.6.0 out tomorrow or this weekend. The new OLPC rpms will follow immediately after that.

[cscott]

uwog: status? we'd like to get update.1 out soonish.

[uwog]

Tomorrow, would that do cscott?

[uwog]

Uploaded a new rpm to dev.laptop.org.

[erikos]

I tested the rpm in joyride-1796 as working.

I will verify in update.1 as well.

@uwog: the process should be to indicate that you tested in joyride and name who will verify the fix then: http://wiki.laptop.org/go/Update.1_process I know it changed a lot of times :)

[bemasc]

Still broken in Update.1-702.

[erikos]

Yup since the libabiword rpm never made it in 702. I hope we get another build.

[erikos]

Tested as fixed in 703.