Printf Formatting Bug

[joshklod]

When using the `%f` specifier of `printf`, if the value is rounded up, the last trailing 0 is dropped.

Tested with:

• ARMbasic[9.37m]
• BASICtools[6.06]
• ARMbasic Kernel[8.36a]
• SuperPRO

Example:

Code: Select all

dim a as single
a = -1
printf("%%1.4f: %1.4f   %%1.2f: %1.2f   %%1.1f: %1.1f", a, a, a)
while a < 2
	a += 1/256
	printf("%%1.4f: %1.4f  %%1.2f: %1.2f  %%1.1f: %1.1f", a, a, a)
loop

Output:

Code: Select all

%1.4f: -1.0000   %1.2f: -1.00   %1.1f: -1.0
%1.4f: -0.9961  %1.2f: -1.0  %1.1f: -1.
%1.4f: -0.9922  %1.2f: -0.99  %1.1f: -1.
%1.4f: -0.9883  %1.2f: -0.99  %1.1f: -1.
%1.4f: -0.9844  %1.2f: -0.98  %1.1f: -1.
%1.4f: -0.9805  %1.2f: -0.98  %1.1f: -1.
%1.4f: -0.9766  %1.2f: -0.98  %1.1f: -1.
%1.4f: -0.9727  %1.2f: -0.97  %1.1f: -1.
%1.4f: -0.9688  %1.2f: -0.97  %1.1f: -1.
%1.4f: -0.9648  %1.2f: -0.96  %1.1f: -1.
%1.4f: -0.9609  %1.2f: -0.96  %1.1f: -1.
%1.4f: -0.9570  %1.2f: -0.96  %1.1f: -1.
%1.4f: -0.9531  %1.2f: -0.95  %1.1f: -1.
%1.4f: -0.9492  %1.2f: -0.95  %1.1f: -0.9
%1.4f: -0.9453  %1.2f: -0.95  %1.1f: -0.9
[Omitted correct lines]
%1.4f: 0.9453  %1.2f: 0.95  %1.1f: 0.9
%1.4f: 0.9492  %1.2f: 0.95  %1.1f: 0.9
%1.4f: 0.9531  %1.2f: 0.95  %1.1f: 1.
%1.4f: 0.9570  %1.2f: 0.96  %1.1f: 1.
%1.4f: 0.9609  %1.2f: 0.96  %1.1f: 1.
%1.4f: 0.9648  %1.2f: 0.96  %1.1f: 1.
%1.4f: 0.9688  %1.2f: 0.97  %1.1f: 1.
%1.4f: 0.9727  %1.2f: 0.97  %1.1f: 1.
%1.4f: 0.9766  %1.2f: 0.98  %1.1f: 1.
%1.4f: 0.9805  %1.2f: 0.98  %1.1f: 1.
%1.4f: 0.9844  %1.2f: 0.98  %1.1f: 1.
%1.4f: 0.9883  %1.2f: 0.99  %1.1f: 1.
%1.4f: 0.9922  %1.2f: 0.99  %1.1f: 1.
%1.4f: 0.9961  %1.2f: 1.0  %1.1f: 1.
%1.4f: 1.0000  %1.2f: 1.00  %1.1f: 1.0
%1.4f: 1.0039  %1.2f: 1.00  %1.1f: 1.0
[Omitted correct lines]
%1.4f: 1.9453  %1.2f: 1.95  %1.1f: 1.9
%1.4f: 1.9492  %1.2f: 1.95  %1.1f: 1.9
%1.4f: 1.9531  %1.2f: 1.95  %1.1f: 2.
%1.4f: 1.9570  %1.2f: 1.96  %1.1f: 2.
%1.4f: 1.9609  %1.2f: 1.96  %1.1f: 2.
%1.4f: 1.9648  %1.2f: 1.96  %1.1f: 2.
%1.4f: 1.9688  %1.2f: 1.97  %1.1f: 2.
%1.4f: 1.9727  %1.2f: 1.97  %1.1f: 2.
%1.4f: 1.9766  %1.2f: 1.98  %1.1f: 2.
%1.4f: 1.9805  %1.2f: 1.98  %1.1f: 2.
%1.4f: 1.9844  %1.2f: 1.98  %1.1f: 2.
%1.4f: 1.9883  %1.2f: 1.99  %1.1f: 2.
%1.4f: 1.9922  %1.2f: 1.99  %1.1f: 2.
%1.4f: 1.9961  %1.2f: 2.0  %1.1f: 2.
%1.4f: 2.0000  %1.2f: 2.00  %1.1f: 2.0

[basicchip]

printf is part of the firmware, and I duplicated your test.

the printf code is open source code we found the web and have been using for some time, without noticing this error. We publish the source of that code in the C tools. Also a link here

https://coridium.us/files/printf_breakpoint.zip

We will look into it, but don't know when we will have a fix, which would require a firmware change.

[basicchip]

I found the issue, not sure why that special case was handled incorrectly in the code. We did not write that originally and I can only assume rounding was disabled for most cases. We enable rounding in BASIC as that seems to be the standard.

We will post a firmware update here, and start rolling that into production products. Here is the test case I used for the fixed version --

Code: Select all

dim x as single
x=0.9996
print x
printf("%1.3f",x)
x=9.9996
print x
printf("%1.3f",x)
printf("%1.3E  %1.4E",x,x)
printf("%1.3G %1.4G",x,x)

Output --

Code: Select all

0.9996000
1.000
9.999600
10.000
1.000E+01  9.9996E+00
10.000 9.9996

[basicchip]

I have fixed the rollover rounding bug.

You need to copy TclTerm.tcl and LPC1756.hex files attached into the directory where BASICtools.exe is run from (typically /Program Files (x86)/Coridium)

The attachment TclTerm.tcl is no longer available

LPC1756..hex

(71.2 KiB) Downloaded 1529 times

[basicchip]

the latest phpBB3 seems to have a bug when trying to attach multiple files, so here is TclTerm.tcl

TclTerm.tcl

(73.35 KiB) Downloaded 1547 times

[joshklod]

The new firmware works perfectly. All my issues have been resolved.

Thank you very much for the fast fix!