Bill Davis (bdavis@NREL.ColoState.EDU) wrote: : Has anyone encountered any major headaches when upgrading Ingres V6.402 from : SUN OS4.3 to Solaris (SUN OS5.4). We will be upgrading our entire system of : Sun SPARC workstations next week and I am interested in hearing about any : unpleasant experiences outside of the normal pains of installing the new : Ingres software or unloading and reloading the database. Hi Bill, we have just done the same operation, and these are the main problems we encountered . . . . - Solaris 5.4 as default does NOT come with a compiler - no compiler no Ingres, period. - If you choose to use the latest gcc (5.? I can't remember) - beware In addition to changes to the Ingres make script (iilink in the utility subdirectory) you need to make changes to various other files for any software development in ABF or W4GL. Also we had numerous problems with double (f8) precision fp in some ESQL/C stuff. It seemed that the compiler was not correctly Word aligning sizeof(double) in some of our SQLDA stuff - leading to erratic Core Dumps. - We gave up on gcc after four days due to the above problem without finding a sensible resolution to it. - If you choose the SUN SparcWorks compiler make sure you get the right version (2.1 or higher) or you could have problems, and as they don't link to /bin/cc as default you will either have to do this manually or put the /opt/SUNWspro (or whatever) path into the chkins script in the utility subdirectory. - We also had problems (this made no sense but believe me it happened) porting executables from one Solaris 2.4 Ingres installation to another. Basically, any ESQL/C program with an EXEC SQL CONNECT TO .. IDENTIFIED AS ... statement compiled on one Installation would not port to another installation where the II_SYSTEM variable was different!!!. For the doubting thomases - I have truss output showing the executable trying to open files in the compiler hosts filesystems!!! The workaround was of course to make sure we compiled on the installation where the software was to be used. - There was also a problem (well weakness) caused by careless coding if you have a big ESQL/C program which connects to the database multiple times, and you submit an EXEC SQL DISCONNECT request when the program is not currently connected to a database (though it has been previously) this causes the program to 'runaway' - ie get caught in a loop. Trussing of the program shows the ingres library functions trapping sigsegv - then causing a sigsegv - which is trapped - causing a sigsegv . . . usw. This one we couldn't fix. Thats all I can think of for the moment (I don't have my notes with me at present). Despite this - it's well worth it - the performance of our dual processor boxes is heaps better. Rgds Mike +=======================================================================+ | ,---. Michael Oakland - Ingres DBA and Technical Support | | / . . \ Consulting to Western Mining Corporation ACN 004 184 598 | | * * Home: (09) 325 1957 MIS: (09) 479 0887 | | \\___// MIS: oakland@wmcmis.dialix.oz.au | | `---' An ORB is for life, not just for Christmas | +=======================================================================+
© William Yuan 2000
Email William