IBM PL/I Refresh 2.1.7 - 09/00

• Enhancements
• Compatibility Improvements
• Fixes

Refresh 2.1.7 Enhancements/Changes

The following enhancements and changes have been added:

• WorkFrame has been significantly changed (in large part so that it can work with FFS).

• The debugger has also been significantly changed (for starters, it is called idebug, not pdebug) and will now work under Windows 2000.

• The Programing Guide, Language Reference Manual and Messages and Codes Book have all been updated with all the changes made up to and including this fixpak.

• Initial support for UTF-16 Unicode is now provided via the new WIDECHAR attribute.

  There is currently no support yet for

  • WIDECHAR characters in source files
  • W string constants
  • use of WIDECHAR expressions in stream I/O
  • implicit conversion to/from WIDECHAR in record I/O
  • implicit endianness flags in record I/O

    If you create a WIDECHAR file, you should write the endianness flag ('fe_ff'wx) as the first two bytes of the file.

• The source listings produced by the macro facility and the compiler are now demarcated by a line saying "MACRO source" and "Compiler source" respectively. The list of macro facility messages is also preceded by a line saying "MACRO messages" (a line saying "Compiler messages" was already produced before the list of compiler messages.)

• If an input file has type crlf, then the library will correctly read its records if they are delimited by either crlf or lf.

• On entry to FROMALIEN routines, the compiler will now save the floating-point control word and then set it to the value needed by PL/I so that floating-point conditions such as zerodivide and underflow can be correctly handled. On exit from such a routine, the saved value will be restored.

  This enables PL/I to work correctly when called from VisualBasic or COBOL.

• Conversions from 4 byte (or smaller) REAL FIXED BIN with a zero scale factor will now be done inline.

• On 390, the DD option has one new suboption. This suboption can be used to specify an alternate DD name for SYSLIB.

  By default, the option is DD(SYSPRINT,SYSIN,SYSLIB,SYSPUNCH,SYSLIB)

• On 390, more conversions are done inline:
  • more fixed bin to character
  • more picture to float

• The FIXEDDEC(29) suboption of the LIMITS compiler option has been dropped.

• Support has been added for the following built-in functions:
  • ACOSF
  • ASINF
  • ATANF
  • CHARVAL
  • COSF
  • EXPF
  • ISIGNED
  • IUNSIGNED
  • LOG10F
  • LOGF
  • ONWCHAR
  • ONWSOURCE
  • SINF
  • TANF
  • WCHAR
  • WCHARVAL
  • WHIGH
  • WIDECHAR
  • WLOW

• The preprocessor has been enhanced to include support for:
  • Arrays in preprocessor procedures
  • WHILE, UNTIL and LOOP keywords in %DO statements
  • %ITERATE statement
  • %LEAVE statement
  • %REPLACE statement
  • %SELECT statement
  • Additional built-in functions:
    • COLLATE
    • COMMENT
    • COMPILEDATE
    • COMPILETIME
    • COPY
    • COUNTER
    • DIMENSION
    • HBOUND
    • INDEX
    • LBOUND
    • LENGTH
    • MACCOL
    • MACLMAR
    • MACRMAR
    • MAX
    • MIN
    • PARMSET
    • QUOTE
    • REPEAT
    • SUBSTR
    • SYSPARM
    • SYSTEM
    • SYSVERSION
    • TRANSLATE
    • VERIFY

Refresh 2.1.7 Compatibility Improvements

The following changes have been made to increase compatibility with the old host compiler:

• Under RULES(LAXDCL), the compiler will now issue an I-level message for each undeclared variable.

• The macro facility will now, by default, uppercase all input outside of quotes and comments. Without this change, the 390 CICS translator would choke on miexdcase code.

  If you want the macro facility to leave your source as you entered it, you can specify the new MACRO suboption CASE(ASIS), i.e. you would invoke the macro facility via PP(MACRO('CASE(ASIS)'))

• The mdeck output will also be more like the old compiler's mdeck in the placement and handling of blanks.

• Float constants such as 1E+40 are valid on hexadecimal float systems, but not on IEEE float systems. On IEEE systems, the compiler used to flag these constants and replace them by HUGE(1s0). It will now flag them with a message saying that the "E" has been replaced by a "D". If they are assigned to a short float, then the compiler will issue a warning message and replace them by HUGE(1s0).

• Under RULES(LAXCTL) all constant extents in descriptor items having the CONTROLLED attribute will be treated as asterisks.

• The severity of the message for an undeclared macro variable has been lowered to that of a W-level message.

Refresh 2.1.7 Fixes

The following compiler/library problems have been fixed since the last fixpak:

• The compiler would abend if a WRITE statement specified in its FROM clause a BASED variable that ended with an AREA.

• If a CONTROLLED array or structure was passed to an ENTRY that declared its arguments but did not specify that they were CONTROLLED, then the called routine could not ALLOCATE or FREE them. This is now possible as long as the DFT(DESCLOCATOR) compiler option is used.

• The COMPLEX attribute was not shown in the ATTRIBUTES listing for COMPLEX PICTURE variables.

• The compiler failed to accept recursively-defined DEFINE STRUCTUREs.

• The compiler incorretcly handled assignments of the form "p = q + r" where p, q, and r were all PIC'(n)9' with n > 8.

• The macro facility generated incorrect code when a macro with keys was invoked with one of keys having an argument list containing consecutive commas.

• The instruction scheduling optimization introduced by the last csd has been turned off since it caused too many new problems.

• Under RULES(LAXCTL) all constant extents in descriptor items having the CONTROLLED attribute will be treated as asterisks.

• Under DFT(NULL370), pointer to offset conversions were handled incorrectly when the source was equal to null().

• Incorrect code was generated for the UNSIGNED built-in function when a precision of 32 was specified as its second argument.