On 11/09/13 01:04, Laszlo Ersek wrote:
> Laszlo Ersek (4):
> i440fx-test: qtest_start() should be paired with qtest_end()
> i440fx-test: give each GTest case its own qtest
> i440fx-test: generate temporary firmware blob
> i440fx-test: verify firmware under 4G and 1M, both -bios and -pflash
>
> tests/i440fx-test.c | 169
> ++++++++++++++++++++++++++++++++++++++++++++++------
> 1 file changed, 152 insertions(+), 17 deletions(-)
Self-NAK
I'll have to send a new version, for at least two reasons:
- The original code (before patch #1) doesn't bother to free "data.bus"
(the retval of qpci_init_pc()). Therefore I assumed this function only
grabbed a (non-counted) reference. This is not the case: the original
code leaks it (although it exits soon after), and my patch #2 doubles
the leak.
- There's something fishy with g_assert() firing in the new test case. I
end up with a hung qemu process, reparented to init:
x86_64-softmmu/qemu-system-x86_64 \
-qtest unix:/tmp/qtest-26926.sock,nowait -qtest-log /dev/null \
-qmp unix:/tmp/qtest-26926.qmp,nowait \
-pidfile /tmp/qtest-26926.pid -machine accel=qtest \
-S -display none -pflash /tmp/fw_blob_553Y5W
I have no idea if this has to do with my unorthodox use of "-S" in
qtest, or if it's a general shortcoming of qtest (ie. aborting before
qtest_end()).
Anyway the series should be reviewable as-is, so I'll wait a bit for
comments.
Thanks!
Laszlo
