IBM PL/I CSD #1 09/98

• Enhancements
• Fixes

CSD #1 Enhancements/Changes

The following enhancements and changes have been added:

• If an ORDINAL is defined without the PRECISION attribute, it will now occupy only as many bytes as needed by its largest value (instead of having it occupy 4 bytes as it did in previous releases). This makes the mapping match that used by the IBM VisualAge C/C++ compiler, and this makes our documentation correct.

  However, this change is incompatible with the code generated by the previous releases of our PL/I compiler. If you have defined an ordinal that you want to have occupy 4 bytes, you can force that by specifying PRECISION(31) (or PREC(31)) in its definition.

• The EXTNAME suboption of the LIMITS option now accepts any value between 7 and 100 inclusive.

• The default for the EXTNAME suboption of the LIMITS option is now 100.

• The default for the NAME suboption of the LIMITS option is now 100.

• The default for the SYSTEM option (under Intel) is now S486.

• The NOIMPRECISE option has been enabled under AIX.

• The attributes BIGENDIAN and LITTLEENDIAN may now be specified. On Intel, they are synonyms for NONNATIVE and NATIVE, respectively.

• A warning message is now issued for any declaration of a SIGNED FIXED BIN variable (or parameter) with a precision that is not one of the values 7, 15 or 31.

  This message would be produced, for instance, if a variable were declared as FIXED BIN(8), when UNSIGNED FIXED BIN(8) was meant. If the address of this variable were taken and used as if it pointed to an UNSIGNED FIXED BIN(8), the code would work on a LittleEndian system such as Intel, but not on a BigEndian system such as AIX or MVS.

• FIXEDOVERFLOW is now raised only for FIXED DECIMAL on both AIX and Intel.

• The macro facility now also allows names to start with an underscore.

• Two new SQL sample programs demonstrating DB2 Stored Procedures have been added (SQSRVOUT.PLI and SQCLIOUT.PLI).

• The SQL Preprocessor now allows variables with the same name to be declared within the same program while adhereing to the PL/I scoping rules.

CSD #1 Fixes

The following problems have been fixed:

• If a STATIC structure contained an UNALIGNED BIT element, the last byte of that element might not be correctly initialized.

• Implicit declares were not flagged if they were for variables in a nested procedure.

• If you built your own ibmwstb.dll, it was honored by the compiler, but ignored, under Windows, by the runtime.

• FILEREAD (and similar built-in functions) would fail if the file to which it applied was declared and opened in a separate procedure.

• The DFT(INITFILL) option had no affect on adjustable automatic variables.

• The compiler would sometimes produce an erroneous error message for compound assignments where the target had asterisk subscripts.

• The compiler would produce an erroneous error message if a default statement specifying the DIMENSION attribute applied to any block that declared a CONTROLLED variable.

• If a program used a simple reference to a bit variable of unknown length as a conditional expression (for instance, if it had a statement of the form "IF x THEN" where "x" was a "BIT(*)" parameter), the compilation would die with an error in the code generation phase.

• If a program contained an ENTRY statement with a RETURNS clause and the DEFAULT(RETURNS(BYADDR)) option was not specified, then the compilation would die with an error in the build phase.

• In a statement of the form "DISPLAY( x ) REPLY( y )" where "y" was CHARACTER VARYING, if a string with length longer than the maximum length of "y" was entered for the REPLY value, the string would be truncated to fit "y", but the length of "y" would be set incorrectly.

• If the a value equal (or near) 1Q-4915 was assigned to a float, an attempt to display it or convert it to character, incorrectly raised overflow.

• If a PICTURE variable containing a generalized dollar, e.g. PIC'999', was converted to FIXED DECIMAL or to another PICTURE type, the wrong results would be produced.

• In assignments of the form A += B, where A and B had the same structuring, the compound assignment operator was incorrectly treated as a simple assignment operator.

• If a variable was declared as DEFINED on a variable that was a constant, no message was issued.

• If a GRAPHIC variable was declared with the VALUE attribute but without a length, then the compilation would die with an error in the build phase.

• If a BASED HANDLE had an INITIAL attribute that specified NULL or SYSNULL as its initial value, any ALLOCATE statement for that variable would be flagged as in error (when actually ok).

• Protection exception was occurring for PUT statements that have the COPY option.

• The SQL Preprocessor now allows the following style of DEFINED declarations for Host compatibility: Dcl Col_Array(11) bin fixed (15); Dcl Col_Array_01 bin fixed (15) DEF Col_Array (01);