# New Ticket Created by Ron Blaschke
# Please include the string: [perl #42254]
# in the subject line of all future correspondence about this issue.
# <URL: http://rt.perl.org/rt3/Ticket/Display.html?id=42254 >
This showed up on Cygwin, but is most likely a problem on other
platforms too. Test t/src/io.t#4 (read) goes into an infinite loop if
F<temp.file> is missing.
io = PIO_open(interp, NULL, "temp.file", "<");
/* this is for testing buffers, not for performance */
PIO_setbuf(interp, io, 4);
do {
len = PIO_read(interp, io, buf, 3);
buf[len] = '\0';
/* dont write trailing spaces */
PIO_printf(interp, "%d: %s\n", len, len ? buf : "EOF");
} while (len > 0);
On Cygwin F<temp.file> is missing because the write test uses a not
exported function and therefore doesn't even compile.
Now, the problem is that the returned C<len> is 16777215 (0x00FFFFFF),
which is larger than zero and keeps the loop going until the "16777215:"
output consumes enough memory (most of the time for quite large values
of "enough").
This can easily be tested by unlink-ing F<temp.file> just before the
read test, and running C<prove -v t/src/io.t>.
I think this bug affects the GCC toolchain, because I see it on Cygwin,
MinGW and Linux (Ubuntu).
Now here's the funny part: All t/src/io.t tests are skipped on Win32
(Visual C++) because they are broken. But if I manually compile the
read test and run it I get:
C:\tmp>io_4.exe
Hello, World!
3: Hel
3: lo,
3: Wo
3: rld
2: !
0: EOF
C:\tmp>del temp.file
C:\tmp>io_4.exe
-1:
Things I guess should be done:
* Add test for missing file
* limit number of iterations in above loop
* S&D bug
