On Mon, Aug 28, 2017 at 09:54:38AM +0200, Łukasz Majewski wrote:
On 08/28/2017 01:02 AM, Chaiken, Alison wrote:
On 2017-05-31 01:12, Lukasz Majewski wrote:
[ . . . ]
And another request -> Could you consider adding tests for those new gpt commands to the 'sandbox' (sandbox_defconfig) ?
Then you can 'mount' some gpt test image ('host' command) and use it with: gpt <command> host X .....
The GPT functionality really cannot be tested without a block device. We could build the disk.raw described in board/sandbox/README.sandbox when (CONFIG_UNIT_TEST && CONFIG_CMD_GPT_RENAME) in imitation of the file helloworld.efi. There would be no reason for the size to be as large as 1200M, of course. 256KB appears to be the smallest usable size for 4 partitions (although trying to partition a 16 KB raw object with gdisk produces entertaining results).
To create this object for testing, we could use a subdirectory like lib/efi_loader, but maybe doing so in test/ makes more sense? If the build system did create the object, then we could modify board/sandbox/README.sandbox to so indicate. Then the test should go in test/py/tests presumably.
If you have encountered some issues with adding this test to sandbox (and those are difficult to solve), then it would be also correct to add this test to test/py and run it on a real HW.
Right. Everyone can run test.py on sandbox _and_ QEMU, and there are then some people with labs setup that also run test.py on real hardware. Between QEMU and real hardware we should be able to add and run tests for this. Thanks!