ARRAY INSERT bug?

[Bern Ertl]

The following works correctly if you change F20-F22 to WORDS.
But it does not work correctly as is......

#COMPILE EXE
#REGISTER NONE
#DIM ALL


TYPE Ten_Quad
   TQ(10) AS QUAD
END TYPE


TYPE My_UDT
   F1       AS STRING * 80
   F2       AS WORD
   F3       AS WORD
   F4       AS WORD
   F5       AS WORD
   F6       AS WORD
   F7       AS QUAD
   F8       AS DWORD
   F9       AS Ten_Quad
   F10      AS DWORD
   F11      AS DWORD
   F12      AS DWORD
   F13      AS DWORD
   F14      AS DWORD
   F15      AS DWORD
   F16      AS DWORD
   F17      AS DWORD
   F18      AS DWORD
   F19      AS WORD
   F20      AS BYTE
   F21      AS BYTE
   F22      AS BYTE
END TYPE


%MYEQUATE    = 202


FUNCTION PBMAIN () AS LONG


   LOCAL SBuff AS My_UDT


   DIM ABuff( 1: 10) AS My_UDT


   SBuff.F1 = "THIS IS INSERTED"
   SBuff.F2 = 8
   SBuff.F19 = 8
   SBuff.F20 = 8
   SBuff.F21 = 8
   SBuff.F22 = %MYEQUATE


   ARRAY INSERT ABuff( 4), SBuff


   MSGBOX "ABuff(4).F1 ="+RTRIM$(ABuff( 4).F1)+ $TAB + "SBuff.F1 ="+RTRIM$(SBuff.F1) + $CRLF + _
          "ABuff(4).F2 ="+STR$(ABuff( 4).F2)+ $TAB + "SBuff.F2 ="+STR$(SBuff.F2) + $CRLF + _
          "ABuff(4).F19 ="+STR$(ABuff( 4).F19)+ $TAB + "SBuff.F19 ="+STR$(SBuff.F19) + $CRLF + _
          "ABuff(4).F20 ="+STR$(ABuff( 4).F20)+ $TAB + "SBuff.F20 ="+STR$(SBuff.F20) + $CRLF + _
          "ABuff(4).F21 ="+STR$(ABuff( 4).F21)+ $TAB + "SBuff.F21 ="+STR$(SBuff.F21) + $CRLF + _
          "ABuff(4).F22 ="+STR$(ABuff( 4).F22)+ $TAB + "SBuff.F22 ="+STR$(SBuff.F22)


END FUNCTION

------------------
Bernard Ertl

[Lance Edmonds]

Thanks, however, this one has been reported here before. R&D assure me that it will definitely be corrected in the next update to the compiler.

To be clear, the bug is described thus "ARRAY INSERT may clear the last byte of an inserted user defined type under certain conditions".

------------------
Lance
PowerBASIC Support
mailto:[email protected][email protected]</A>

[Eric Pearson]

Bern --

You can also add an extra byte to your structure (like F23 AS BYTE) to work around the problem. I'm not sure, but it looks to me like "data alignment" is involved.

BTW, even after the compiler glitch is fixed, there are some good reasons to keep your structures WORD aligned or, better yet, DWORD aligned. 32-bit processors can perform their tasks more efficiently when data is aligned on 32-bit boundaries. The CPU actually has to perform extra steps to access 1- and 2- byte numeric values. That's why LONG variables are the fastest numeric type, with DWORD a close second.

-- Eric


------------------
Perfect Sync Development Tools
Perfect Sync Web Site
Contact Us: mailto:[email protected][email protected]</A>

[Guest]

Does the TYPE "alignment" parameter also fix the problem by automagically
padding the structure as necessary to achieve the specified alignment.

TYPE My_UDT
  F2       AS WORD
  F3       AS WORD
  F4       AS BYTE
END TYPE  '  Might have problem ???
 
TYPE My_UDT DWORD
  F2       AS WORD
  F3       AS WORD
  F4       AS BYTE
END TYPE  '  WILL NOT have problem ???

[This message has been edited by Ron Pierce (edited March 05, 2001).]

[Eric Pearson]

Ron --

It seems to. I tried WORD and DWORD alignment and Bern's problem disappeared.

But remember, I am just guessing (at this point) that alignment is the root of the problem. It could be that we are just masking the real problem, and that a different structure will behave differently. Until we get a confirmation from PB, I would not consider this to be a reliable workaround.

-- Eric

------------------
Perfect Sync Development Tools
Perfect Sync Web Site
Contact Us: mailto:[email protected][email protected]</A>

[Lance Edmonds]

If the problem is actually evident, the very last byte (and only the last byte) of the inserted UDT structure may be affected.

Therefore, simply adding a dummy byte to pad the structure, or using any alignment that effectively pads the UDT to provide an additional byte should be a sufficient workaround.

Remember, the total byte count of a UDT is affected by the alignment method specified. For example, if DWORD is used, then the UDT will be a multiple of 4 bytes in length, etc.

Therefore, a structure such as...

TYPE MyData DWORD
  V1 AS INTEGER
  V2 AS INTEGER
  V3 AS INTEGER
  V4 AS INTEGER
  V5 AS STRING * 5
END TYPE

...will be 24 bytes in length.




------------------
Lance
PowerBASIC Support
mailto:[email protected][email protected]</A>

[Guest]

Lance, doesn't byte alignment only pad the total length of a structure/UDT to
a size evenly divisible by the size of the alignment data type?

For instance, in your example:
[code]
TYPE MyData DWORD
V1 AS INTEGER
V2 AS INTEGER
V3 AS INTEGER
V4 AS INTEGER
V5 AS STRING * 5
END TYPE
... I would have expected the length of the structure to be padded to 16 bytes
instead of each variable in the structure being padded to a DWORD boundary.

Why is the structure length padded to 24 bytes instead of 16 bytes? Is each variable in
the structure padded to DWORD alignment instead of the structure
being padded to a DWORD alignment size?

[Tom Hanlin]

The alignment refers to the members of the TYPE, e.g., WORD alignment
will cause each member of the UDT to be word-aligned. This is typical
behavior for every compiler I've seen that supports custom alignment.

------------------
Tom Hanlin
PowerBASIC Staff