On 03/09/2015 12:27 PM, Kevin Wolf wrote:
Am 09.03.2015 um 17:03 hat John Snow geschrieben:
On 03/09/2015 10:27 AM, Kevin Wolf wrote:
Am 26.02.2015 um 00:06 hat John Snow geschrieben:
+void mkimg(const char *file, const char *fmt, unsigned size_mb)
+{
+ gchar *cli;
+ bool ret;
+ int rc;
+ GError *err = NULL;
+ char *qemu_img_path;
+ gchar *out, *out2;
+
+ qemu_img_path = getenv("QTEST_QEMU_IMG");
+ assert(qemu_img_path);
+
+ cli = g_strdup_printf("./%s create -f %s %s %uM", qemu_img_path,
+ fmt, file, size_mb);
+ ret = g_spawn_command_line_sync(cli, &out, &out2, &rc, &err);
+ if (err) {
+ fprintf(stderr, "%s\n", err->message);
+ g_error_free(err);
+ }
+ g_assert(ret && !err);
+
+ ret = g_spawn_check_exit_status(rc, &err);
This function only exists since glib 2.34. Dropping the following
patches from the queue:
pick 23134a5 qtest/ahci: add qcow2 support to ahci-test
pick 6ca5609 qtest/ahci: test different disk sectors
pick e2f0dee qtest/ahci: Add simple flush test
pick 396491b qtest/ahci: Allow override of default CLI options
pick eb8c8bd libqtest: add qmp_eventwait
pick 398bfc3 libqtest: add qmp_async
pick d3f77d1 libqos: add blkdebug_prepare_script
pick d628e51 qtest/ahci: add flush retry test
This is patch 7 and 8 from this series and the complete series "ahci:
rerror/werror=stop resume tests", which seems to depend on them.
Kevin
Do we have a policy on glib version for qtests? I know we require a
specific version for within QEMU itself, but there are many
instances of functions newer than that being using in qtests
already.
I would have assumed that it's the same requirement and make check
should work whereever you can build qemu. In practice, however, I'm
running qtests on a RHEL 6 host (without this series that still works
fine) and complaining only if things break there.
Kevin
It looks like we currently require 2.12 for the basic builds, but pretty
much the entirety of the gtest framework itself was introduced in 2.16,
hence the ask.
I saw some discussion recently of bumping the glib version required, but
I think MST circumvented the need to do so for the time-being.
Still, we should at least bump this up to 2.16 to have some consistency
with the testing framework... Or acknowledge that the testing framework
may have a different set of requirements.
--js
