Sorry about forgetting that. I've been somewhat focused on the windows
port today.
2006-08-07 Andrew Black <[EMAIL PROTECTED]>
* output.cpp (check_example): Initialize read buffers prior to read to
prevent incorrect results
Martin Sebor wrote:
Andrew Black wrote:
Greetings all.
Attached is a short patch that should resolve this issue.
Yup, looks good, thanks!
Can you please post a ChangeLog to go with your change so I can
commit it?
Thanks again!
Martin
The cause of the error was that the output was smaller than the static
holding buffer, so the call to memcmp was comparing beyond the amount
of data that had been read into the buffer. This in turn was leading
to a difference in the unfilled portions of the buffer. This
difference was caused by the natural variations in uninitialized memory.
--Andrew Black
Martin Sebor wrote:
There are a bunch of examples that the exec utility reports as
having failed with the OUTPUT status even though when executed
by hand they run successfully to completion and produce the
expected output. I suspect there must be a bug in the utility.
$ make auto_ptr run; ./auto_ptr > auto_ptr.out && diff auto_ptr.out
/build/sebor/stdcxx/examples/manual/out/auto_ptr.out; echo $?
make: `auto_ptr' is up to date.
NAME STATUS ASSERTS FAILED PERCNT
auto_ptr OUTPUT
0
Martin
------------------------------------------------------------------------
Index: output.cpp
===================================================================
--- output.cpp (revision 429465)
+++ output.cpp (working copy)
@@ -296,6 +296,10 @@
size_t out_read, ref_read;
int match = 1;
+ /* Zero out holding buffers to avoid false differences */
+ memset (out_buf, 0, DELTA_BUF_LEN);
+ memset (ref_buf, 0, DELTA_BUF_LEN);
+
while (!feof (reference) && !feof (output)) {
/* First, read a block from the files into the buffer */
out_read = fread (out_buf, DELTA_BUF_LEN, 1, output);
