forums.silverfrost.com

[Anonymous]

Hi there! This is my first post ... so be gentle!

I use FTN95 for Net and Win32 (but still at version 4.0.0 ... this may be relevant!).

I have a project with a number of functions and subroutines together with a single data module (the regular stuff!). When everything is together as a single programme, it works fine ... no compiler errors or warnings. So, I thought, let's make it into a .DLL structure.

So, I carefully split out all functions and subroutines into individual units, compiled them with the command:

ftn95 /FULL_UNDEF /ZEROISE /FIXED_FORMAT %1 /MOD_PATH ..MODULES

Then combined all .OBJ files into a .DLL with a .lnk file containing the commands:

dll
lo *.obj
subsystem windows
exportall
file mydll

All seems to work as expected!

Then I run a test programme compiled with the commands:

ftn95 /ZEROISE /FIXED_FORMAT %1 /MOD_PATH MODULES

and link in the files with the commands:

echo subsystem windows > inttemp$.lnk
echo load %1 >> inttemp$.lnk
echo load mydll.dll >> inttemp$.lnk
echo load MODULESmydata_DATA >> inttemp$.lnk
echo file >> inttemp$.lnk
slink inttemp$.lnk

Everything works until I run the programme ... at which point I get an ACCESS VIOLATION initialising a matrix in the data module. This was not a problem in the full programme ... nor is it a problem if I build another test programme that simply initialises that matrix in the data module on its own - but using the mydll.dll. Very puzzling!

Has anyone any thoughts on the solution --- is it the compiler version? Is it a compiler option? Is it dirty programming that only becomes obvious in the .DLL environment.

All thoughts and comments very welcome. Thank you in anticipation.


Norm