[PATCH 0/4] freebsd portability fixes

I've been setting up auto-builders for a few platforms to try to catch
portability problems early. Unfortunately, it was made harder by the
fact that the _current_ test scripts don't pass on all platforms.

With these patches, I can successfully run all test scripts from the
current 'master' on FreeBSD 6.1 (some of them probably affect OS X, too
-- I recall somebody complaining about the '! foo | bar' construct
recently).

-Peff

On some shells (notably /bin/sh on FreeBSD 6.1), the
construct

  ! foo | bar

does not negate the exit value of the pipeline, but rather
of 'foo', producing the opposite of the expected value. We
can work around this by specifying it as

  ! (foo | bar)

Signed-off-by: Jeff King <redacted>
---
 git-rebase.sh     |    2 +-
 t/t3400-rebase.sh |    4 ++--
 t/t3700-add.sh    |    6 +++---
 t/t7501-commit.sh |    2 +-
 4 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/git-rebase.sh b/git-rebase.sh
index 9b13b83..fbb0f28 100755
--- a/git-rebase.sh
+++ b/git-rebase.sh

@@ -353,7 +353,7 @@ orig_head=$branch
 mb=$(git merge-base "$onto" "$branch")
 if test "$upstream" = "$onto" && test "$mb" = "$onto" &&
 	# linear history?
-	! git rev-list --parents "$onto".."$branch" | grep " .* " > /dev/null
+	! (git rev-list --parents "$onto".."$branch" | grep " .* ") > /dev/null
 then
 	# Lazily switch to the target branch if needed...
 	test -z "$switch_to" || git checkout "$switch_to"

diff --git a/t/t3400-rebase.sh b/t/t3400-rebase.sh
index 496f4ec..fdad7da 100755
--- a/t/t3400-rebase.sh
+++ b/t/t3400-rebase.sh

@@ -44,13 +44,13 @@ test_expect_success 'rebase against master' '
 
 test_expect_success \
     'the rebase operation should not have destroyed author information' \
-    '! git log | grep "Author:" | grep "<>"'
+    '! (git log | grep "Author:" | grep "<>")'
 
 test_expect_success 'rebase after merge master' '
      git reset --hard topic &&
      git merge master &&
      git rebase master &&
-     ! git show | grep "^Merge:"
+     ! (git show | grep "^Merge:")
 '
 
 test_expect_success 'rebase of history with merges is linearized' '

diff --git a/t/t3700-add.sh b/t/t3700-add.sh
index 287e058..68c5dde 100755
--- a/t/t3700-add.sh
+++ b/t/t3700-add.sh

@@ -81,17 +81,17 @@ test_expect_success '.gitignore test setup' '
 
 test_expect_success '.gitignore is honored' '
 	git add . &&
-	! git ls-files | grep "\\.ig"
+	! (git ls-files | grep "\\.ig")
 '
 
 test_expect_success 'error out when attempting to add ignored ones without -f' '
 	! git add a.?? &&
-	! git ls-files | grep "\\.ig"
+	! (git ls-files | grep "\\.ig")
 '
 
 test_expect_success 'error out when attempting to add ignored ones without -f' '
 	! git add d.?? &&
-	! git ls-files | grep "\\.ig"
+	! (git ls-files | grep "\\.ig")
 '
 
 test_expect_success 'add ignored ones with -f' '

diff --git a/t/t7501-commit.sh b/t/t7501-commit.sh
index c0288f3..89710af 100755
--- a/t/t7501-commit.sh
+++ b/t/t7501-commit.sh

@@ -41,7 +41,7 @@ test_expect_success \
 test_expect_success \
 	"using paths with --interactive" \
 	"echo bong-o-bong >file &&
-	! echo 7 | git-commit -m foo --interactive file"
+	! (echo 7 | git-commit -m foo --interactive file)"
 
 test_expect_success \
 	"using invalid commit with -C" \

-- 
1.5.5.1.296.gf618c

The output of 'tar tv' varies from system to system. In
particular, the t5000 was expecting to parse the date from
something like:

  -rw-rw-r-- root/root         0 2008-05-13 04:27 file

but FreeBSD's tar produces this:

  -rw-rw-r--  0 root   root        0 May 13 04:27 file

Instead of relying on tar's output, let's just extract the
file using tar and stat the result using perl.

Signed-off-by: Jeff King <redacted>
---
 t/t5000-tar-tree.sh |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/t/t5000-tar-tree.sh b/t/t5000-tar-tree.sh
index fa62b6a..9b0baac 100755
--- a/t/t5000-tar-tree.sh
+++ b/t/t5000-tar-tree.sh

@@ -67,10 +67,10 @@ test_expect_success \
 
 test_expect_success \
     'validate file modification time' \
-    'TZ=GMT $TAR tvf b.tar a/a |
-     awk \{print\ \$4,\ \(length\(\$5\)\<7\)\ ?\ \$5\":00\"\ :\ \$5\} \
-     >b.mtime &&
-     echo "2005-05-27 22:00:00" >expected.mtime &&
+    'mkdir extract &&
+     $TAR xf b.tar -C extract a/a &&
+     perl -e '\''print((stat("extract/a/a"))[9], "\n")'\'' >b.mtime &&
+     echo "1117231200" >expected.mtime &&
      diff expected.mtime b.mtime'
 
 test_expect_success \

-- 
1.5.5.1.296.gf618c

When using /bin/sh from FreeBSD 6.1, the value of $? is lost
when calling a function inside the 'trap' action. This
resulted in clone erroneously indicating success when it
should have reported failure.

As a workaround, we save the value of $? before calling any
functions.

Signed-off-by: Jeff King <redacted>
---
 git-clone.sh |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/git-clone.sh b/git-clone.sh
index 9d88d1c..547228e 100755
--- a/git-clone.sh
+++ b/git-clone.sh

@@ -240,7 +240,6 @@ die "working tree '$GIT_WORK_TREE' already exists."
 D=
 W=
 cleanup() {
-	err=$?
 	test -z "$D" && rm -rf "$dir"
 	test -z "$W" && test -n "$GIT_WORK_TREE" && rm -rf "$GIT_WORK_TREE"
 	cd ..

@@ -248,7 +247,7 @@ cleanup() {
 	test -n "$W" && rm -rf "$W"
 	exit $err
 }
-trap cleanup 0
+trap 'err=$?; cleanup' 0
 mkdir -p "$dir" && D=$(cd "$dir" && pwd) || usage
 test -n "$GIT_WORK_TREE" && mkdir -p "$GIT_WORK_TREE" &&
 W=$(cd "$GIT_WORK_TREE" && pwd) && GIT_WORK_TREE="$W" && export GIT_WORK_TREE

-- 
1.5.5.1.296.gf618c

filter-branch tries to restore "old" copies of some
environment variables by using the construct:

  unset var
  test -z "$old_var" || var="$old_var" && export var

However, by the short-circuit logic, we will always run
'export var'. On bash and dash, exporting an unset variable
has no effect. However, on some shells (such as FreeBSD's
/bin/sh), the shell exports the empty value.

This manifested itself in this case as git-filter-branch
setting GIT_INDEX_FILE to the empty string, which in turn
caused its call to git-read-tree to fail, leaving the
working tree pointing at the original HEAD instead of the
rewritten one.

To fix this, we change the short-circuit logic to better
match the intent:

  test -z "$old_var" || { var="$old_var" && export var; }

Signed-off-by: Jeff King <redacted>
---
 git-filter-branch.sh |   10 +++++-----
 1 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/git-filter-branch.sh b/git-filter-branch.sh
index 333f6a8..0304dc5 100755
--- a/git-filter-branch.sh
+++ b/git-filter-branch.sh

@@ -435,11 +435,11 @@ rm -rf "$tempdir"
 trap - 0
 
 unset GIT_DIR GIT_WORK_TREE GIT_INDEX_FILE
-test -z "$ORIG_GIT_DIR" || GIT_DIR="$ORIG_GIT_DIR" && export GIT_DIR
-test -z "$ORIG_GIT_WORK_TREE" || GIT_WORK_TREE="$ORIG_GIT_WORK_TREE" &&
-	export GIT_WORK_TREE
-test -z "$ORIG_GIT_INDEX_FILE" || GIT_INDEX_FILE="$ORIG_GIT_INDEX_FILE" &&
-	export GIT_INDEX_FILE
+test -z "$ORIG_GIT_DIR" || { GIT_DIR="$ORIG_GIT_DIR" && export GIT_DIR; }
+test -z "$ORIG_GIT_WORK_TREE" || { GIT_WORK_TREE="$ORIG_GIT_WORK_TREE" &&
+	export GIT_WORK_TREE; }
+test -z "$ORIG_GIT_INDEX_FILE" || { GIT_INDEX_FILE="$ORIG_GIT_INDEX_FILE" &&
+	export GIT_INDEX_FILE; }
 git read-tree -u -m HEAD
 
 exit $ret

-- 
1.5.5.1.296.gf618c

On Tue, May 13, 2008 at 04:43:39AM -0400, Jeff King wrote:

With these patches, I can successfully run all test scripts from the
current 'master' on FreeBSD 6.1 (some of them probably affect OS X, too
-- I recall somebody complaining about the '! foo | bar' construct
recently).

Ah, nevermind about OS X. It was 97ad535b from Alex, and he specifically
mentioned FreeBSD 4.

Alex, you might want to try re-running the tests with these patches.

-Peff

Jeff King, Tue, May 13, 2008 11:04:24 +0200:

On Tue, May 13, 2008 at 04:43:39AM -0400, Jeff King wrote:

quoted

With these patches, I can successfully run all test scripts from the
current 'master' on FreeBSD 6.1 (some of them probably affect OS X, too
-- I recall somebody complaining about the '! foo | bar' construct
recently).

Ah, nevermind about OS X. It was 97ad535b from Alex, and he specifically
mentioned FreeBSD 4.

Alex, you might want to try re-running the tests with these patches.

Very good! GnuPG is missing on the system, so t7004 was skipped,
but everything besides that ran fine. My config.mak:

NO_STRTOUMAX=Yes
NO_C99_FORMAT=Yes
SHELL_PATH=/usr/local/bin/bash

perl 5.8.6

I also have the test-tr patches I sent some time ago in that tree.

Alex Riesen, Tue, May 13, 2008 22:39:31 +0200:

Jeff King, Tue, May 13, 2008 11:04:24 +0200:

quoted

On Tue, May 13, 2008 at 04:43:39AM -0400, Jeff King wrote:

quoted

With these patches, I can successfully run all test scripts from the
current 'master' on FreeBSD 6.1 (some of them probably affect OS X, too
-- I recall somebody complaining about the '! foo | bar' construct
recently).

Ah, nevermind about OS X. It was 97ad535b from Alex, and he specifically
mentioned FreeBSD 4.

Alex, you might want to try re-running the tests with these patches.

Very good! GnuPG is missing on the system, so t7004 was skipped,
but everything besides that ran fine. My config.mak:

NO_STRTOUMAX=Yes
NO_C99_FORMAT=Yes
SHELL_PATH=/usr/local/bin/bash

perl 5.8.6

I also have the test-tr patches I sent some time ago in that tree.

I also have the following in git-compat-util.h:

diff --git a/git-compat-util.h b/git-compat-util.h
index 01c4045..161bd50 100644
--- a/git-compat-util.h
+++ b/git-compat-util.h

@@ -191,6 +191,7 @@ extern size_t gitstrlcpy(char *, const char *, size_t);
 
 #ifdef NO_STRTOUMAX
 #define strtoumax gitstrtoumax
+typedef unsigned long long uintmax_t;
 extern uintmax_t gitstrtoumax(const char *, char **, int);
 #endif
 

Which is obviously a hack, but I considered too minor an issue. It
(the missing uintmax_t) is probably something very specific to this
particular system. I am probably wrong about the issue being minor...