[avrdude-dev] Minor efficiency question

[Larry Colen]

As I posted to the list a couple of days ago, I'm trying to figure out
why perfectly legitimate hex files are causing avrdude to give me 
errors.  In the process of trying to understand what the relevant code 
is doing I found what seems to be an error in the code, though it's more 
of an inefficiency.

At about line 175 in fileio.c, in ihex_readrec(), there is a test at the 
beginning that there are at least two characters available for the byte 
count:

  if (offset + 2 > len)              
    return -1;

The way I read the description of the intel hex format the minimum legal 
length of a record is 11 bytes.  It seems to me that if the function 
were just to test that the record length were at least 9, once it was 
computed, you could drop several of the intermediate checks that the 
record was long enough to make the next calculation.

//  :ccAAAA0tDDDDDDDDss
//   cc                  Byte count
//     AAAA              Address
//         0t            Record Type
//           DDDDDDDD    Data
//                   ss  Checksum

If each case where the test for length could fail, a different error 
code were returned, I could see some diagnostic use for the multiple 
tests, but since they all just return -1, it seems that you could save a 
few CPU cycles for each record by just making sure that the length were 
greater than the minimum, and then that the length and the cc value 
agreed with each other.

  Larry

--
Larry Colen address@hidden (from dos4est)