Re: Error -1235 (misleading message)

In article <2E6586C13341E60D.DF49B756DA06A631.F9B3D91E02C0C761@library- proxy.airnews.net>, Daniel Wright <dw420@airmail.net> writes >Well, this is one of those misleading error messages I suppose.... > >My program is written in 4GL and ESQL/C and this is probably totally >unrelated to why it started core-dumping (I must be doing something >wrong), but it consistently core-dumps when I call retflo() with a value >other than zero....no matter if it is hard-coded, cast to a float or >whatever.....but I'm still debugging that. This message is about the >misleading error message I got. > >After adding "WHENEVER ANY ERROR STOP" to the code, I got the following >output (even after I reduced it to a minimum amount of code, with no >database references and no pointer references). > >dev/home/dannyw/dsc1/inv/ictl> ./a.out >Program stopped at "fu_4gl.4gl", line number 8. >FORMS statement error number -1235. >Character host variable is too short for the data. >dev/home/dannyw/dsc1/inv/ictl> finderr 1235 >-1235 Character host variable is too short for the data. > You've tried to put something into a varable and the variable is too small. >In an ESQL/C program, the program has attempted to fetch a column value >into a host variable that is not large enough. Use the DESCRIBE command >to find out the sizes of column values. > >If this error arises in a 4GL program, please note all circumstances, >and contact the Informix Technical Support Department. > >Since I naively believed Informix's explanation and wasn't doing any >FETCH in my EC module, I presumed the 4GL code had caused the problem. >I was all ready to call Informix Tech Support, when I noticed that if I >commented out the popquote() function in my EC, the error went away. > >I didn't particularly care about how my ptr-to-char was popped off the >stack (with a null-terminator or not) since I only intended to look at >the initial character anyway [if (*f_ptr_to_string == 'T') .... else >.....]. > >Incidentally f_ptr_to_string was declared as char f_ptr_to_string[2]; > >I decided to review the manual and change the popquote( f_ptr_to_string, >1); to popquote( f_ptr_to_string, 2 ); > Remember C needs the additional null terminator. 4GL DEFINE c CHAR(1) C char c[2]; /* Include null terminator*/ >Wow, it worked....[except I still core dump, which I guess is a separate >problem]. > >I'm tired of bitching about stupid problems with 4GL which Informix >doesn't care to correct...they've already expressed their disinterest in >maintaining and improving this product, so for what it's worth, don't >believe your error messages. > It's not an Informix problem, your code is wrong.. >The only thing I can imagine is causing my program to core-dump on a >retflo() call is that I must be inadvertently assigning something to a >pointer which I set incorrectly and am thus wiping out the memory where >this function resides. Well you are trashing memory. Remember in which order you return arguments to Informix ( try c4gl -keep and look at the .ec file or check the manals). Also don't forget if you are returning 4 arguments you C code should do a ... return(4); } > >If you can think of another possibility please let me know.....If it >matters, this is on HP-UX 10.20 w/ Informix 7.21.UC1 w/ 4GL 6.04.UC3. I >also got the same -1235 error on an AIX something or other running >Informix 5.something or other.....didn't follow thru to see if the Then it's your code. try compiling with -g and do sdb a.out core $t $c to get a strack trace and then you can tell where the code dump is happening. >entire program gave me the same core dump...too much trouble to port the >database and such, when I think the fault must lie with the C >code....even though myself and a couple of others have reviewed most of >it and found nothing suspect in it.... > Try sending it to us... 4GL + C, >Thanks... > >-- >- Danny Wright >thgirwnad@moc.toofgib (Yeah, I know you have to strain to read that, >but maybe I'll cut down on some spam. Here's a clue, it's >bass-ackwards) >#include <stddisclaimers.h> > > -- David Williams