| Age | Commit message (Collapse) | Author |
|---|
|
if the connection is killed it blanks the pointer it points to, to make it
easier to detect usage problems whereever Curl_done() is used.
|
|
|
|
This is necessary in case the client never connects or somehow fails to do
it timely. The timeout used now is only 2 seconds, which might cause problems
on really slow hosts but longer times are painful when doing torture testing
on FTP test cases.
I'm not sure how this 'alarm' functionality works on Windows or other systems
that don't actually have the alarm() function.
|
|
|
|
memory leak detection
|
|
some (SSL-)options.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
we now use it to provide more info back on connect failures.
|
|
|
|
|
|
|
|
|
|
|
|
Also made -t imply -n to disable valgrind, it runs sloooow otherwise.
This now manages to run all tests OK up to test case 100 (the first FTP one)
for me.
|
|
system thinks free(NULL) is badness)
|
|
|
|
|
|
|
|
|
|
take a CURLcode as a second argument, that is non-zero when Curl_done()
is called after an error was returned from Curl_do() (or similar).
|
|
|
|
to deliver
|
|
|
|
|
|
instead of using snprintf() + strdup().
|
|
defined (which it never is atm).
Now, we can focus on making 'runtests -t [num]' work on all test cases and
we should never leak nor crash.
|
|
|
|
|
|
|
|
|
|
we typecast on invoke.
|
|
causes xlc and gcc act differently on AIX.
|
