Mikael Ronström
2008-May-16 18:47 UTC
[dtrace-discuss] Fwd: DTrace bug with 6 parameters to C++ function containing multiple DTrace probe entries
Hi, Had to forward to another email box to be able to send it :(> > > Hi, > I don''t know how to file bugs towards DTrace yet, but here seems to > be one anyways. > I made a DTrace probe that was part of an inline function, this > inline function was > called in numberous places. However one of those places seems to > cause special pain. > > The function contained 6 parameters and DTrace''s name of it was > Undefined first referenced > symbol in file > _ZN4Dbtc20deleteFromIndexTableEP6SignalPNS_18TcFiredTriggerDataEP3PtrI > NS_16ApiConnectRecordEEPS4_INS_15TcConnectRecordEEPNS_11T blocks/ > ndb_blocks_dtrace.o > The function looked like this: > void deleteFromIndexTable(Signal* signal, > TcFiredTriggerData* firedTriggerData, > ApiConnectRecordPtr* transPtr, > TcConnectRecordPtr* opPtr, > TcIndexData* indexData, > bool holdOperation = false); > > From what I can see this is a simple matter of an array being > declared to short in DTrace since it seems to > cut the name short here. > > The workaround is of course to get rid of a few parameters in some > way and this works, going down to > four parameters and it works fine. However it''s a similar function > where this works so is properly a bitmore complex.> > > Rgrds Mikael > > >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/dtrace-discuss/attachments/20080516/f2e3ab11/attachment.html>
Mikael Ronström
2008-May-16 19:32 UTC
[dtrace-discuss] Fwd: DTrace bug with 6 parameters to C++ function containing multiple DTrace probe entries
Hi, After making a workaround in removing the last attribute and shortening the name of the probe I found that unfortunately now the entire provider is gone when I do dtrace -l which is a tad unfortunate :( Rgrds Mikael 16 maj 2008 kl. 20.47 skrev Mikael Ronstr?m:> Hi, > Had to forward to another email box to be able to send it :( > >> >> >> Hi, >> I don''t know how to file bugs towards DTrace yet, but here seems >> to be one anyways. >> I made a DTrace probe that was part of an inline function, this >> inline function was >> called in numberous places. However one of those places seems to >> cause special pain. >> >> The function contained 6 parameters and DTrace''s name of it was >> Undefined first referenced >> symbol in file >> _ZN4Dbtc20deleteFromIndexTableEP6SignalPNS_18TcFiredTriggerDataEP3Ptr >> INS_16ApiConnectRecordEEPS4_INS_15TcConnectRecordEEPNS_11T blocks/ >> ndb_blocks_dtrace.o >> The function looked like this: >> void deleteFromIndexTable(Signal* signal, >> TcFiredTriggerData* firedTriggerData, >> ApiConnectRecordPtr* transPtr, >> TcConnectRecordPtr* opPtr, >> TcIndexData* indexData, >> bool holdOperation = false); >> >> From what I can see this is a simple matter of an array being >> declared to short in DTrace since it seems to >> cut the name short here. >> >> The workaround is of course to get rid of a few parameters in some >> way and this works, going down to >> four parameters and it works fine. However it''s a similar function >> where this works so is properly a bit > more complex. >> >> >> Rgrds Mikael >> >> >> > > _______________________________________________ > dtrace-discuss mailing list > dtrace-discuss at opensolaris.org-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/dtrace-discuss/attachments/20080516/4e91e1d3/attachment.html>
Adam Leventhal
2008-May-16 21:46 UTC
[dtrace-discuss] Fwd: DTrace bug with 6 parameters to C++ function containing multiple DTrace probe entries
Hi Mikael, You can file bugs at bugs.opensolaris.org. The information you''ve provided isn''t really sufficient to be able to diagnose the problem (for example, what provider are you talking about). Please provide as much relevant detail as you can, and any relevant output from dtrace(1M). Adam On Fri, May 16, 2008 at 09:32:37PM +0200, Mikael Ronstr?m wrote:> Hi, > After making a workaround in removing the last attribute and > shortening the name of the probe I found that unfortunately > now the entire provider is gone when I do dtrace -l which is > a tad unfortunate :( > > Rgrds Mikael > > 16 maj 2008 kl. 20.47 skrev Mikael Ronstr?m: > >> Hi, >> Had to forward to another email box to be able to send it :( >> >>> >>> >>> Hi, >>> I don''t know how to file bugs towards DTrace yet, but here seems to be >>> one anyways. >>> I made a DTrace probe that was part of an inline function, this inline >>> function was >>> called in numberous places. However one of those places seems to cause >>> special pain. >>> >>> The function contained 6 parameters and DTrace''s name of it was >>> Undefined first referenced >>> symbol in file >>> _ZN4Dbtc20deleteFromIndexTableEP6SignalPNS_18TcFiredTriggerDataEP3PtrINS_16ApiConnectRecordEEPS4_INS_15TcConnectRecordEEPNS_11T >>> blocks/ndb_blocks_dtrace.o >>> The function looked like this: >>> void deleteFromIndexTable(Signal* signal, >>> TcFiredTriggerData* firedTriggerData, >>> ApiConnectRecordPtr* transPtr, >>> TcConnectRecordPtr* opPtr, >>> TcIndexData* indexData, >>> bool holdOperation = false); >>> >>> From what I can see this is a simple matter of an array being declared to >>> short in DTrace since it seems to >>> cut the name short here. >>> >>> The workaround is of course to get rid of a few parameters in some way >>> and this works, going down to >>> four parameters and it works fine. However it''s a similar function where >>> this works so is properly a bit >> more complex. >>> >>> >>> Rgrds Mikael >>> >>> >>> >> >> _______________________________________________ >> dtrace-discuss mailing list >> dtrace-discuss at opensolaris.org >> _______________________________________________ > dtrace-discuss mailing list > dtrace-discuss at opensolaris.org-- Adam Leventhal, Fishworks http://blogs.sun.com/ahl
Mikael Ronström
2008-May-16 22:55 UTC
[dtrace-discuss] Fwd: DTrace bug with 6 parameters to C++ function containing multiple DTrace probe entries
Hi Adam, Ok, I''ll file a bug a bit later, easiest manner I gather is to simply upload the non-functional MySQL source tarball with a description of how to build it and run it. Rgrds Mikael 16 maj 2008 kl. 23.46 skrev Adam Leventhal:> Hi Mikael, > > You can file bugs at bugs.opensolaris.org. The information you''ve > provided > isn''t really sufficient to be able to diagnose the problem (for > example, > what provider are you talking about). Please provide as much > relevant detail > as you can, and any relevant output from dtrace(1M). > > Adam > > On Fri, May 16, 2008 at 09:32:37PM +0200, Mikael Ronstr?m wrote: >> Hi, >> After making a workaround in removing the last attribute and >> shortening the name of the probe I found that unfortunately >> now the entire provider is gone when I do dtrace -l which is >> a tad unfortunate :( >> >> Rgrds Mikael >> >> 16 maj 2008 kl. 20.47 skrev Mikael Ronstr?m: >> >>> Hi, >>> Had to forward to another email box to be able to send it :( >>> >>>> >>>> >>>> Hi, >>>> I don''t know how to file bugs towards DTrace yet, but here seems >>>> to be >>>> one anyways. >>>> I made a DTrace probe that was part of an inline function, this >>>> inline >>>> function was >>>> called in numberous places. However one of those places seems to >>>> cause >>>> special pain. >>>> >>>> The function contained 6 parameters and DTrace''s name of it was >>>> Undefined first referenced >>>> symbol in file >>>> _ZN4Dbtc20deleteFromIndexTableEP6SignalPNS_18TcFiredTriggerDataEP3P >>>> trINS_16ApiConnectRecordEEPS4_INS_15TcConnectRecordEEPNS_11T >>>> blocks/ndb_blocks_dtrace.o >>>> The function looked like this: >>>> void deleteFromIndexTable(Signal* signal, >>>> TcFiredTriggerData* firedTriggerData, >>>> ApiConnectRecordPtr* transPtr, >>>> TcConnectRecordPtr* opPtr, >>>> TcIndexData* indexData, >>>> bool holdOperation = false); >>>> >>>> From what I can see this is a simple matter of an array being >>>> declared to >>>> short in DTrace since it seems to >>>> cut the name short here. >>>> >>>> The workaround is of course to get rid of a few parameters in >>>> some way >>>> and this works, going down to >>>> four parameters and it works fine. However it''s a similar >>>> function where >>>> this works so is properly a bit >>> more complex. >>>> >>>> >>>> Rgrds Mikael >>>> >>>> >>>> >>> >>> _______________________________________________ >>> dtrace-discuss mailing list >>> dtrace-discuss at opensolaris.org >> > >> _______________________________________________ >> dtrace-discuss mailing list >> dtrace-discuss at opensolaris.org > > > -- > Adam Leventhal, Fishworks http:// > blogs.sun.com/ahl