combo-layer: avoid too long command lines in update with history

As suspected, invoking "git archive" with all intended files as
parameters can run into command line length limitations. Splitting up
the parameters into multiple invocations (xargs-style) works and was
tested after encountering the situation in practice.

Signed-off-by: Patrick Ohly <patrick.ohly@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

1 files changed, 29 insertions, 2 deletions

diff --git a/scripts/combo-layer b/scripts/combo-layer
index a0a737d6ee..1ca2ce6c02 100755
--- a/scripts/combo-layer
+++ b/scripts/combo-layer
@@ -1266,8 +1266,35 @@ def apply_commit(parent, rev, largs, wargs, dest_dir, file_filter=None):
         target = os.path.join(wargs["destdir"], dest_dir)
         if not os.path.isdir(target):
             os.makedirs(target)
-        runcmd("git archive %s %s | tar -C %s -xf -" % (rev, ' '.join([pipes.quote(x) for x in update]), pipes.quote(target)), **largs)
-        runcmd("git add -f".split() + [os.path.join(dest_dir, x) for x in update], **wargs)
+        quoted_target = pipes.quote(target)
+        # os.sysconf('SC_ARG_MAX') is lying: running a command with
+        # string length 629343 already failed with "Argument list too
+        # long" although SC_ARG_MAX = 2097152. "man execve" explains
+        # the limitations, but those are pretty complicated. So here
+        # we just hard-code a fixed value which is more likely to work.
+        max_cmdsize = 64 * 1024
+        while update:
+            quoted_args = []
+            unquoted_args = []
+            cmdsize = 100 + len(quoted_target)
+            while update:
+                quoted_next = pipes.quote(update[0])
+                size_next = len(quoted_next) + len(dest_dir) + 1
+                logger.debug('cmdline length %d + %d < %d?' % (cmdsize, size_next, os.sysconf('SC_ARG_MAX')))
+                if cmdsize + size_next < max_cmdsize:
+                    quoted_args.append(quoted_next)
+                    unquoted_args.append(update.pop(0))
+                    cmdsize += size_next
+                else:
+                    logger.debug('Breaking the cmdline at length %d' % cmdsize)
+                    break
+            logger.debug('Final cmdline length %d / %d' % (cmdsize, os.sysconf('SC_ARG_MAX')))
+            cmd = "git archive %s %s | tar -C %s -xf -" % (rev, ' '.join(quoted_args), quoted_target)
+            logger.debug('First cmdline length %d' % len(cmd))
+            runcmd(cmd, **largs)
+            cmd = "git add -f".split() + [os.path.join(dest_dir, x) for x in unquoted_args]
+            logger.debug('Second cmdline length %d' % reduce(lambda x, y: x + len(y), cmd, 0))
+            runcmd(cmd, **wargs)
     if delete:
         for path in delete:
             if dest_dir: