By definition, anything that the compiler handles is source code so this 529,000 lines whether or not it is data initialization is source code. I can't even imagine the difficulty is fixing one or two values in something like that. This almost has to be generated code. I would seriously look at loading this data at run-time. In the case of very large static data sets, one technique I have used which greatly reduces the initization time required to load large chunks of data at run-time is to initialize the data block and store it in shared memory. This is a one-time run and then subsequent invokations of the same program examine shared memory to see if the data is already loaded. If so, they simply attach. The idea is that the first load might take many minutes but subsequent startups of other processes bypass this step.
Even so, I haven't found any hard limits that releate to the number of lines of code. There are hard limits for things like the maximum identifier length, maximum number of dimensions of an array. I also notice that this is 32-bit code so you may have exceeded a barrier that would not exist in 64-bit compiles. When a process receives a SIGSEGV (11), the error message may not really describe the event.
You should really examine the compiler options (e.g +Onolimit) to see if anything looks as though it might help but since you didn't bother to mention compiler version, it's difficult to look for specific options or patches.
If it ain't broke, I can fix that.
