I have a DataSnap Win32 service application built with AnyDAC running on SQL 2008 R2 that appears to be experiencing memory corruption issues.
I do not have a reproducible test case at the moment, but the evidence seems to point to a problem with TADStoredProc parameter handling in AnyDAC v6.0.5.2837. I reverted back to AnyDAC v5.0.7 which we use in a very similar DataSnap server and have not seen any signs of instability since.
The application issues we are seeing show up as sporadic AVs every day or two. The server is currently under very light load so considerable time elapses between crashes. The MadExcept traces generated by the server show that the failure is occurring at the start of DFM resource parsing for the DataSnap service data module when the client connects. Once the AV occurs, the server is destabilized and subsequent requests crash with an AV in the same place on each new client request. This strongly suggests that memory corruption is occurring.
Today we observed an ugly design-time crash in RadStudio with v6.0.5.2837 in response to changing the stored procedure parameters in the database. When the TADStoredProc pointing to the MSSQL proc with modified parameters was enabled, a BDS exception window was displayed with the message [AnyDAC][Stan][Eval]-107 followed by a bunch of garbage characters (see attached screenshot). The exception is being raised in TADExpressionParser.NextToken. The thought is that this IDE exception may be related to the memory corruption issue we have been seeing.
I will continue to investigate this issue as time allows and will let you know if I can create a reproducible test case or identify the root cause. The main purpose of posting this report is to raise awareness of a potential issue in AnyDAC in case anyone is experiencing a similar problem.
David
http://www.da-soft.com/images/fbfiles/images/AnyDAC_BDS_Crash.jpg
↧