--maxsleep broken for values less than 1000

[GrantEdwards]

On Linux, the --maxsleep option only seems to work correctly for values of 1000 or larger. [Correction, it's broken for any value that isn't a multiple of 1000.] Values smaller than 1000 send packets way faster than they should. It's hard to get accurate test results, but it sort of looks like values less than 1000 are getting treated as microseconds rather than milliseconds. A value of 999 sends around 1000pps, and a value of 500 sends about 2000pps. I get pretty much the same same results for -T gtod, nano, select.

[GrantEdwards]

To convert from ms to ns, you need to multiply by 1000000 instead of 1000:

diff --git a/src/tcpreplay_api.c b/src/tcpreplay_api.c
index fde27d6a..b6124358 100644
--- a/src/tcpreplay_api.c
+++ b/src/tcpreplay_api.c
@@ -200,7 +200,7 @@ tcpreplay_post_args(tcpreplay_t *ctx, int argc)
 
     if (HAVE_OPT(MAXSLEEP)) {
         options->maxsleep.tv_sec = OPT_VALUE_MAXSLEEP / 1000;
-        options->maxsleep.tv_nsec = (OPT_VALUE_MAXSLEEP % 1000) * 1000;
+        options->maxsleep.tv_nsec = (OPT_VALUE_MAXSLEEP % 1000) * 1000000;
     }

[fklassen]

Thanks for the bug report and the patch. The patch fixes the issue.

[fklassen]

Fixed in Pr #460