From 1f5f147a9ec048897c43e2358223430098493497 Mon Sep 17 00:00:00 2001
From: Rob Landley <rob@landley.net>
Date: Thu, 24 Oct 2019 19:41:09 -0500
Subject: Fix the xargs argument too long problem by putting the proper
 accounting back when they haven't specified -s, add tests.

---
 tests/xargs.test   |  9 +++++++++
 toys/posix/xargs.c | 12 +++++-------
 2 files changed, 14 insertions(+), 7 deletions(-)

diff --git a/tests/xargs.test b/tests/xargs.test
index e2eec036..afed8a17 100644
--- a/tests/xargs.test
+++ b/tests/xargs.test
@@ -52,6 +52,15 @@ testing "false" "xargs false; echo \$?" "123\n" "" ""
 testing "exit 255" "xargs sh -c 'exit 255' 2>&1; echo \$?" \
   "xargs: sh: exited with status 255; aborting\n124\n" "" ""
 
+# findutils maxes out at 131066 (they do the math wrong), kernel takes 131071.
+toyonly testing "max argument size" \
+  "head -c 131071 /dev/zero | tr '\0' x | xargs | wc -c" "131072\n" "" ""
+X=$(head -c 131066 /dev/zero | tr '\0' x)
+# This goes over the kernel's 10 meg limit
+testing "big input forces split" \
+  'for i in $(seq 1 100);do echo $X;done|{ xargs >/dev/null && echo yes;}' \
+  "yes\n" "" ""
+
 # TODO: what exactly is -x supposed to do? why does coreutils output "one"?
 #testing "-x" "xargs -x -s 9 || echo expected" "one\nexpected\n" "" "one\ntwo\nthree"
 
diff --git a/toys/posix/xargs.c b/toys/posix/xargs.c
index a8ecd4d6..725eb593 100644
--- a/toys/posix/xargs.c
+++ b/toys/posix/xargs.c
@@ -64,11 +64,10 @@ static char *handle_entries(char *data, char **entry)
       if (!*s) break;
       save = ss = s;
 
-      // We ought to add sizeof(char *) to TT.bytes to be correct, but we don't
-      // for bug compatibility with busybox 1.30.1 and findutils 4.7.0.
-
+      // Specifying -s can cause "argument too long" errors.
+      if (!FLAG(s)) TT.bytes += sizeof(void *)+1;
       for (;;) {
-        if (++TT.bytes >= TT.s && TT.s) return save;
+        if (++TT.bytes >= TT.s) return save;
         if (!*s || isspace(*s)) break;
         s++;
       }
@@ -102,8 +101,7 @@ void xargs_main(void)
   // that the invoked utility has room to modify its environment variables
   // and command line arguments and still be able to invoke another utility",
   // though obviously that's not really something you can guarantee.
-  bytes = sysconf(_SC_ARG_MAX) - environ_bytes() - 2048;
-  if (!TT.s || TT.s > bytes) TT.s = bytes;
+  if (!FLAG(s)) TT.s = sysconf(_SC_ARG_MAX) - environ_bytes() - 2048;
 
   TT.delim = '\n'*!FLAG(0);
 
@@ -116,7 +114,7 @@ void xargs_main(void)
 
   // count entries
   for (entries = 0, bytes = -1; entries < toys.optc; entries++, bytes++)
-    bytes += strlen(toys.optargs[entries]);
+    bytes += strlen(toys.optargs[entries]) + sizeof(char *)*!FLAG(s);
   if (bytes >= TT.s) error_exit("argument too long");
 
   // Loop through exec chunks.
-- 
cgit v1.2.3