Andrew Black wrote:
Martin Sebor wrote:
[snip]
Hmm, that sure looks like a bug. It would be nice to have a test
case although I wonder if the input required to trigger the endless
loop might need to be invalid?
Could you please open an issue in Jira for this?
Martin
If you want a test case on Solaris, I suspect any of the @euro or UTF-8
locales will show this, among others. These locales are terminating
with a return code of 143 in automated testing, likely due to being
killed by the exec utility.
I tried [EMAIL PROTECTED] with both ISO-8859-1 and UTF-8 and they both
completed so that's not it. The case statement does appear to
handle tok_end (i.e., the END keyword) correctly, and the only
way I can imagine the function reaching the EOF (i.e.,
tok_end_tokens) before it is if the input file were incomplete.
And in fact, this does show the problem:
$ echo "LC_COLLATE" > collate.src \
&& echo > empty.cm \
&& ../bin/localedef -c -w -f empty.cm -i collate.src collate.empty
Martin
