From 1422b7b43a132c786d17935aa4f78620bea5ad87 Mon Sep 17 00:00:00 2001
From: Ole Tange <ole@tange.dk>
Date: Wed, 25 Apr 2018 22:38:42 +0200
Subject: [PATCH] parallel: Fixed bug #53748: -k --retries 10 + out of
 filehandles = blocking.

---
 src/parallel                                 | 13 ++++++++++++-
 testsuite/tests-to-run/parallel-local-10s.sh |  7 +++++++
 testsuite/wanted-results/parallel-local-10s  |  9 +++++++++
 3 files changed, 28 insertions(+), 1 deletion(-)

diff --git a/src/parallel b/src/parallel
index c45b6d4d..49297417 100755
--- a/src/parallel
+++ b/src/parallel
@@ -2729,6 +2729,7 @@ sub init_run_jobs {
 	    $no_more_file_handles_warned++ or
 		::warning("No more file handles. ",
 			  "Raising ulimit -n or /etc/security/limits.conf may help.");
+	    debug("start", "No more file handles. ");
 	    return 0;
 	}
     }
@@ -3973,7 +3974,9 @@ sub reaper {
         }
         my $sshlogin = $job->sshlogin();
         $sshlogin->dec_jobs_running();
-        if(not $job->should_be_retried()) {
+        if($job->should_be_retried()) {
+	    $job->free_ressources();
+	} else {
 	    # The job is done
 	    $sshlogin->inc_jobs_completed();
 	    # Free the jobslot
@@ -9126,6 +9129,14 @@ sub print_tag {
     return print_normal(@_);
 }
 
+sub free_ressources() {
+    my $self = shift;
+    for my $fdno (sort { $a <=> $b } keys %Global::fd) {
+	close $self->fh($fdno,"w");
+	close $self->fh($fdno,"r");
+    }
+}
+
 sub print_normal {
     my $self = shift;
     my ($fdno,$in_fh,$out_fd) = @_;
diff --git a/testsuite/tests-to-run/parallel-local-10s.sh b/testsuite/tests-to-run/parallel-local-10s.sh
index 65b7bb2e..d6d3d0fd 100644
--- a/testsuite/tests-to-run/parallel-local-10s.sh
+++ b/testsuite/tests-to-run/parallel-local-10s.sh
@@ -299,6 +299,13 @@ par_linebuffer_tag_slow_output() {
     parallel --delay 1 -j0 --tag --line-buffer halfline ::: a b
 }
 
+par_retries_all_fail() {
+    echo "bug #53748: -k --retries 10 + out of filehandles = blocking"
+    ulimit -n 30
+    seq 8 |
+	parallel -k -j0 --retries 2 --timeout 0.1 'echo {}; sleep {}; false' 2>/dev/null
+}
+
 export -f $(compgen -A function | grep par_)
 compgen -A function | grep par_ | sort |
     parallel --joblog /tmp/jl-`basename $0` -j10 --tag -k '{} 2>&1'
diff --git a/testsuite/wanted-results/parallel-local-10s b/testsuite/wanted-results/parallel-local-10s
index 5ccee16a..77f6e43f 100644
--- a/testsuite/wanted-results/parallel-local-10s
+++ b/testsuite/wanted-results/parallel-local-10s
@@ -506,6 +506,15 @@ par_results_csv	   	8,:,999.999,999.999,0,6,0,0,"echo 23 11",23,11,"23 11
 par_results_csv	   	",
 par_results_csv	   	9,:,999.999,999.999,0,6,0,0,"echo 23 12",23,12,"23 12
 par_results_csv	   	",
+par_retries_all_fail	bug #53748: -k --retries 10 + out of filehandles = blocking
+par_retries_all_fail	1
+par_retries_all_fail	2
+par_retries_all_fail	3
+par_retries_all_fail	4
+par_retries_all_fail	5
+par_retries_all_fail	6
+par_retries_all_fail	7
+par_retries_all_fail	8
 par_round_robin_blocks	bug #49664: --round-robin does not complete
 par_round_robin_blocks	8
 par_slow_total_jobs	bug #51006: Slow total_jobs() eats job