[PATCH v2] iotests/028: Fix for long $TEST_DIRs

[PATCH v2] iotests/028: Fix for long $TEST_DIRs

[Max Reitz]

For long test image paths, the order of the "Formatting" line and the
"(qemu)" prompt after a drive_backup HMP command may be reversed.  In
fact, the interaction between the prompt and the line may lead to the
"Formatting" to being greppable at all after "read"-ing it (if the
prompt injects an IFS character into the "Formatting" string).

So just wait until we get a prompt.  At that point, the block job must
have been started, so "info block-jobs" will only return "No active
jobs" once it is done.

Reported-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
v2:
- Fix another kind of race...
---
 tests/qemu-iotests/028     | 11 ++++++++---
 tests/qemu-iotests/028.out |  1 -
 2 files changed, 8 insertions(+), 4 deletions(-)

diff --git a/tests/qemu-iotests/028 b/tests/qemu-iotests/028
index 71301ec6e5..bba1ee59ae 100755
--- a/tests/qemu-iotests/028
+++ b/tests/qemu-iotests/028
@@ -119,9 +119,14 @@ fi
 # Silence output since it contains the disk image path and QEMU's readline
 # character echoing makes it very hard to filter the output. Plus, there
 # is no telling how many times the command will repeat before succeeding.
-_send_qemu_cmd $h "drive_backup disk ${TEST_IMG}.copy" "(qemu)" >/dev/null
-_send_qemu_cmd $h "" "Formatting" | _filter_img_create
-qemu_cmd_repeat=20 _send_qemu_cmd $h "info block-jobs" "No active jobs" >/dev/null
+# (Note that creating the image results in a "Formatting..." message over
+# stdout, which is the same channel the monitor uses.  We cannot reliably
+# wait for it because the monitor output may interact with it in such a
+# way that _timed_wait_for cannot read it.  However, once the block job is
+# done, we know that the "Formatting..." message must have appeared
+# already, so the output is still deterministic.)
+silent=y _send_qemu_cmd $h "drive_backup disk ${TEST_IMG}.copy" "(qemu)"
+silent=y qemu_cmd_repeat=20 _send_qemu_cmd $h "info block-jobs" "No active jobs"
 _send_qemu_cmd $h "info block-jobs" "No active jobs"
 _send_qemu_cmd $h 'quit' ""
 
diff --git a/tests/qemu-iotests/028.out b/tests/qemu-iotests/028.out
index 7d54aeb003..37aed84436 100644
--- a/tests/qemu-iotests/028.out
+++ b/tests/qemu-iotests/028.out
@@ -468,7 +468,6 @@ No errors were found on the image.
 
 block-backup
 
-Formatting 'TEST_DIR/t.IMGFMT.copy', fmt=IMGFMT size=4294968832 backing_file=TEST_DIR/t.IMGFMT.base backing_fmt=IMGFMT
 (qemu) info block-jobs
 No active jobs
 === IO: pattern 195
-- 
2.21.0

Re: [PATCH v2] iotests/028: Fix for long $TEST_DIRs

[Thomas Huth]

On 11/10/2019 14.18, Max Reitz wrote:
> For long test image paths, the order of the "Formatting" line and the
> "(qemu)" prompt after a drive_backup HMP command may be reversed.  In
> fact, the interaction between the prompt and the line may lead to the
> "Formatting" to being greppable at all after "read"-ing it (if the
> prompt injects an IFS character into the "Formatting" string).
> 
> So just wait until we get a prompt.  At that point, the block job must
> have been started, so "info block-jobs" will only return "No active
> jobs" once it is done.
> 
> Reported-by: Thomas Huth <thuth@redhat.com>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
> v2:
> - Fix another kind of race...

Thanks, that fixes the issue for me!

Tested-by: Thomas Huth <thuth@redhat.com>

Re: [PATCH v2] iotests/028: Fix for long $TEST_DIRs

[John Snow]

On 10/11/19 8:18 AM, Max Reitz wrote:
> For long test image paths, the order of the "Formatting" line and the
> "(qemu)" prompt after a drive_backup HMP command may be reversed.  In
> fact, the interaction between the prompt and the line may lead to the
> "Formatting" to being greppable at all after "read"-ing it (if the
> prompt injects an IFS character into the "Formatting" string).
> 
> So just wait until we get a prompt.  At that point, the block job must
> have been started, so "info block-jobs" will only return "No active
> jobs" once it is done.
> 
> Reported-by: Thomas Huth <thuth@redhat.com>
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
> v2:
> - Fix another kind of race...
> ---
>  tests/qemu-iotests/028     | 11 ++++++++---
>  tests/qemu-iotests/028.out |  1 -
>  2 files changed, 8 insertions(+), 4 deletions(-)
> 
> diff --git a/tests/qemu-iotests/028 b/tests/qemu-iotests/028
> index 71301ec6e5..bba1ee59ae 100755
> --- a/tests/qemu-iotests/028
> +++ b/tests/qemu-iotests/028
> @@ -119,9 +119,14 @@ fi
>  # Silence output since it contains the disk image path and QEMU's readline
>  # character echoing makes it very hard to filter the output. Plus, there
>  # is no telling how many times the command will repeat before succeeding.
> -_send_qemu_cmd $h "drive_backup disk ${TEST_IMG}.copy" "(qemu)" >/dev/null
> -_send_qemu_cmd $h "" "Formatting" | _filter_img_create
> -qemu_cmd_repeat=20 _send_qemu_cmd $h "info block-jobs" "No active jobs" >/dev/null
> +# (Note that creating the image results in a "Formatting..." message over
> +# stdout, which is the same channel the monitor uses.  We cannot reliably
> +# wait for it because the monitor output may interact with it in such a
> +# way that _timed_wait_for cannot read it.  However, once the block job is
> +# done, we know that the "Formatting..." message must have appeared
> +# already, so the output is still deterministic.)
> +silent=y _send_qemu_cmd $h "drive_backup disk ${TEST_IMG}.copy" "(qemu)"
> +silent=y qemu_cmd_repeat=20 _send_qemu_cmd $h "info block-jobs" "No active jobs"
>  _send_qemu_cmd $h "info block-jobs" "No active jobs"
>  _send_qemu_cmd $h 'quit' ""
>  
> diff --git a/tests/qemu-iotests/028.out b/tests/qemu-iotests/028.out
> index 7d54aeb003..37aed84436 100644
> --- a/tests/qemu-iotests/028.out
> +++ b/tests/qemu-iotests/028.out
> @@ -468,7 +468,6 @@ No errors were found on the image.
>  
>  block-backup
>  
> -Formatting 'TEST_DIR/t.IMGFMT.copy', fmt=IMGFMT size=4294968832 backing_file=TEST_DIR/t.IMGFMT.base backing_fmt=IMGFMT
>  (qemu) info block-jobs
>  No active jobs
>  === IO: pattern 195
> 

Reviewed-by: John Snow <jsnow@redhat.com>

Re: [PATCH v2] iotests/028: Fix for long $TEST_DIRs

[Kevin Wolf]

Am 11.10.2019 um 14:18 hat Max Reitz geschrieben:
> For long test image paths, the order of the "Formatting" line and the
> "(qemu)" prompt after a drive_backup HMP command may be reversed.  In
> fact, the interaction between the prompt and the line may lead to the
> "Formatting" to being greppable at all after "read"-ing it (if the
> prompt injects an IFS character into the "Formatting" string).
> 
> So just wait until we get a prompt.  At that point, the block job must
> have been started, so "info block-jobs" will only return "No active
> jobs" once it is done.
> 
> Reported-by: Thomas Huth <thuth@redhat.com>
> Signed-off-by: Max Reitz <mreitz@redhat.com>

Thanks, applied to the block branch.

Kevin