parallel: Fix to really avoid running new jobs when load is too high.

The conditions were ORed before, which typically made the load limit be
ignored since the jobs limit was satisfied.

Also, -l 0 makes little sense, so don't really wait for the load to become
lower than 0 in that case.

diff --git a/debian/changelog b/debian/changelog
index b5d2a7f1e7675dba0dded49535719ccd2327f4fe..fe89fdfaa25d77dd6e947f7fe6824bbdc6790d18 100644 (file)
--- a/debian/changelog
+++ b/debian/changelog
@@ -3,6 +3,7 @@ moreutils (0.39) UNRELEASED; urgency=low
   * Cap sillyness. Closes: #570815
   * parallel: Fix exit code handling when commands are specified after --
   * parallel: Make -j 0 do something reasonable (start all jobs at once).
+  * parallel: Fix to really avoid running new jobs when load is too high.
 
  -- Joey Hess <joeyh@debian.org>  Sun, 21 Feb 2010 13:16:10 -0500
 

diff --git a/parallel.c b/parallel.c
index 06f9041da58355106579797cc945e136f1d85da5..19f341aa92a42696d419ea0f7a4591bf16176077 100644 (file)
--- a/parallel.c
+++ b/parallel.c
@@ -201,8 +201,9 @@ int main(int argc, char **argv) {
 
                getloadavg(&load, 1);
 
-               if ((maxjobs == 0 || curjobs < maxjobs) ||
-                   (maxload > 0 && load < maxload)) {
+               if ((maxjobs == 0 || curjobs < maxjobs) &&
+                   (maxload <= 0 || load < maxload)) {
+
                        if (argsatonce > arglen - argidx)
                                argsatonce = arglen - argidx;
                        exec_child(command, arguments + argidx,