Recent changes to bugs

Query Result Spec parser fails

David Lopez — Mon, 24 Sep 2007 11:28:43 -0000

For this example query result request: "_user.ID,.Dele.ID,.Nom,.Cog,._uiprof"; as documented, it should interpret that we want the ID value for User.Dele field (instead of the whole record referenced by User.Dele). The binary parser fails, as creates a new object tree (with oID==0) for that field. It should be present into parent object request definition instead. This is the generated tree:
Test_0x1010(): drawing query object decomposition
+-- 1 '_User'(oRec oID=1) bSz=884 NumF=1 Flg=0x0000
| `-- 'id' cc=0x00(=x) cc2=0x01 rID=0 NumV=1 fSz=2
* Type 0
Test_0x1010(): drawing query object result request decomposition
+-- 0 '_User'(oRec oID=1) bSz=320 NumF=4 Flg=0x00 nti=0, nRef=0
| |-- 'usr'
| |-- 'nom' -->Noms
| |-- 'cog'
| `-- '_uiprof' -->_UIap
`-- 2 ''(oRec oID=0) bSz=320 NumF=1 Flg=0x00 nti=0, nRef=1 {oID=1}
`-- 'dele'
As a guide, result should be like when we perform this other example result query specification: "_user.ID,.Dele,.Nom,.Cog,._uiprof", that generates the following tree:
+-- 1 '_User'(oRec oID=1) bSz=884 NumF=1 Flg=0x0000
| `-- 'id' cc=0x00(=x) cc2=0x01 rID=0 NumV=1 fSz=2
* Type 0
Test_0x1010(): drawing query object result request decomposition
`-- 0 '_User'(oRec oID=1) bSz=320 NumF=5 Flg=0x00 nti=0, nRef=0
|-- 'usr'
|-- 'dele' -->Empr
|-- 'nom' -->Noms
|-- 'cog'
`-- '_uiprof' -->_UIap
When we request User.Dele.ID, line with 'dele' -->Empr should change to just 'dele' (with no reference), instead of creating a new branch, etc...

Log system bug when rotating files

David Lopez — Thu, 08 Sep 2005 10:53:14 -0000

There's a misconception into the kinlog.c source file: that
this is the only process to access a file descriptor. It works
ok just for writting, as the OS serializes the writes, but it
fails to keep track of the size of the file, as it only checks
the first time, when it is opened, and takes care of the
written bytes, but the ones witten by current instance, not
by others!! So when the file grows, and it is rotated, some
instances write to the old file as for them, it is still small and
already opened...
The solution has to come without performance penalties,
maybe using a shared semaphore area for the files size,
etc. A more complex approach could be to pipe all the logs
to a specialized process, but this adds some overheat
(slight, but...)

Various SEGV in module IM (this is multiple bugs actually)

David Lopez — Fri, 17 Jun 2005 19:46:33 -0000

Have to trace SAS allocation, etc. for first block,
something leaks there...
BTW, have to investigate also this:
From debug.log: ---->
20050617-214121.001813 (5738) I
korec_ObjectDataBlockRecv: Block size mismatch (local
is 4, storage is 32756) for _WCIdd
<-----

Sockets close suddenly

David Lopez — Thu, 16 Jun 2005 19:22:14 -0000

It seems that when a third node module connects to QM,
connections to previous node are closed somehow.
Sequence is:
1. QM is launched
2. BM is launched, it connects with QM and everything
goes right
3. IM or TM is launched. When it connects to QM, BM
drops connection to QM, then QM also drops connection
to IM/TM... so failure seems to come from QM

From debug.log ------>
20050616-211355.085415 (4283) T kc_SyncCon: sent Sync
to 192.168.0.3:4222 (44 bytes)
20050616-211355.085429 (4283) T kincon_Flush:
connection established to 192.168.0.3:4222
20050616-211355.085452 (4283) T kintsk_Proc: executing
Tsk 1 (core_boot_BT). wa=50, wb=2, wc=1, wd=0
20050616-211355.085464 (4283) T kc_ReqTopology: No
connection to any QM (yet?)
20050616-211355.085502 (4283) T kincon_Flush: sent 44
bytes to 192.168.0.3:4222
20050616-211355.438045 (4268) Q kintsk_Event: sending
event wdTO to Tsk 1 (core_boot_BT), 4 bytes
20050616-211355.438123 (4268) Q
kincore_ScanListenSocks: Have a new TCP (Packet) connection
20050616-211355.438152 (4268) Q kincon_GetNewData: read
44 bytes from 192.168.0.3:37466
20050616-211355.438218 (4268) Q kincon_EndCon: deleted
connection 1 (to 192.168.0.3:37465)
20050616-211355.438220 (4277) B kincon_GetNewData:
error? on socket 5 (read() returns 0, poll status is
0005). RxB_Free=4096
20050616-211355.438238 (4268) Q kincon_EndCon: deleted
connection 0 (to 192.168.0.3:37465)
20050616-211355.438256 (4277) B kincon_GetNewData: no
error could be checked
20050616-211355.438258 (4268) Q kincon_EndCon: freed a
whole connection management block
20050616-211355.438286 (4268) Q kintsk_Proc: executing
Tsk 1 (core_boot_BT). wa=1002, wb=0, wc=3, wd=0
20050616-211355.438308 (4268) Q kintsk_Proc: executing
Tsk 1 (core_boot_BT). wa=1000, wb=0, wc=3, wd=0
20050616-211355.438344 (4277) B kincon_DelCon: entity 1
is not valid anymore
20050616-211355.438363 (4277) B kincon_DelCon: deleted
connection 1 (to 192.168.0.3:4222)
20050616-211355.438374 (4277) B kincon_DelCon: freed a
whole connection management block
<------