View previous topic :: View next topic |
Author |
Message |
silverfrost Site Admin
Joined: 29 Nov 2006 Posts: 191 Location: Manchester
|
Posted: Tue Dec 31, 2019 5:35 pm Post subject: Silverfrost FTN95: Fortran for Windows version 8.60 |
|
|
We are pleased to announce the release of FTN95 version 8.60. This is available for download now to supported users.
New Features:
- Runtime checking has been extended so that FTN95 currently passes all tests in the Polyhedron diagnostic test suite. This includes a new option /CHECK_ALIAS that checks for overlapping arguments in a call to a routine. See ftn95.chm for further details.
- Compile time argument typing checking has been added for routines in clearwin.ins. This will be extended to the corresponding MODULEs in a later release.
- Compile time argument type checking has been added for certain FTN95 intrinsics such as UPCASE@.
- 64 bit executables can now have a stack size that is greater than 4GB. This feature was incorrectly announced for v8.50.
- /NO_TRUNCATE has been implemented for free format files.
- ClearWin+: A pipe symbol '|' may now be used in the help string for a menu item that has a grey control
- ClearWin+: The option [tooltips] has been added for a statusbar (%sb)
- ClearWin+: The option [strings] has been added for a %mb toolbar. The effect is to replace what otherwise would be tooltips by text below each button bitmap
- ClearWin+: On an experimental basis, a %pv pivot can now be used without %ww
- ClearWin+: A new routine SUPPRESS_CMD_BOX@ has been added
- ClearWin+: Improvements have been made regarding the set of routines that where historically based on CREATE_POLYGON@
- ClearWin+: User data (%ud) can now be attached to a window as well as to any control within the window
- ClearWin+: IMPORT_RSC_IMAGE@ has been added as an alternative name for IMPORT_IMAGE@ and a new function IMPORT_FILE_IMAGE@ has been added
- ClearWin+: A new standard callback HELPLOOKUP has been added as an alternative to HELP_LOOKUP
- Microsoft Visual Studio 2019 Plugin
|
|
Back to top |
|
|
silverfrost Site Admin
Joined: 29 Nov 2006 Posts: 191 Location: Manchester
|
Posted: Tue Dec 31, 2019 5:37 pm Post subject: |
|
|
Outline summary of some of the bugs that have been fixed:
- 64 bit SLEEP1@ was not implemented correctly.
- When linking very large executables SLINK could sometimes fail with 'error 42'.
- When trying to use a generic class from .NET the compiler would fail to correctly parse the name
- Compiling code that used INCLUDE statements could cause invalid line number information to be put in the executable. (x64 only)
- Array constructor failure for SUM with array section on left hand side of the assignment.
- Failure in an assignment when calling SIZE for an array of derived TYPE that has an array as a member.
- Various bugs relating to ALLOCATABLE members of a derived TYPE.
- Pointer assignment to NULL() failed with /CHECK.
- Internal write error goes undetected.
- Call to LEN as an argument gives a false warning.
- Failure to fault certain non-conformant array assignments.
- Initialisation of REAL variable in MODULE not propagated (64 bit only).
- Compilation error in CONTAINed subroutines sometimes not detected.
- Scoping error for INTERFACE defined within a MODULE.
- FLOOR and CEILING not working for INTEGER*8 and /64.
- Failure to report pointer assignment error in TYPE.
- Access violation with a certain SAVEd array.
- Various failures relating to certain functions returning character arrays.
- Failure when concatenating certain PARAMETER substrings.
- No warning given when initialising an INTEGER to a CHARACTER.
- Failure with assigned GOTO and /SAVE.
- Failure with a certain call to DOT_PRODUCT (64 bit only).
- CASE(1::3) etc. give an internal compiler error.
- Use of KIND(x) on its own gives a false warning.
- /OPT gives an internal compiler error when declaring a very large array.
- /OPT regression when attempting to unroll a DO loop.
- /OPT failure for 32 bit "erf" functions.
- Two bugs revealed by MODULE EQUIVALENCE involving implicitly declared variable.
- Zero width F FORMAT descriptor was adding a leading space.
|
|
Back to top |
|
|
|