need longer timeout with send_feature_report() for libusb under Linux

[elfmimi]

Some devices may require much longer timeout when submitting hid_send_feature_report() .

for such devices hid_send_feature_report() just notifies failure with return value of -1 .

I checked that internal error code libusb is reporting is LIBUSB_ERROR_TIMEOUT ( -7 ) .

Changing value here, from 1000 to, say, 5000 does make difference.

hidapi/libusb/hid.c

Line 1471 in c3a6150

1000/*timeout millis*/);

You curious what the device is?
It is USB-HID control capable LCD monitor from EIZO . ex. FlexScan EV2670
Most requests to the monitor, for example a request to change speaker volume, work fine with short timeout .
But one specific request that is to enable or disable picture-by-picture mode only works with long timeout .

So it is better to have a measure to somehow specify custom value for timeout.

[Youw]

I guess it was only a matter of time when that magic number would backfire at us.
On the other hand it's been years and no one complained, until now.

But we definitely need a workaround for that. I think a libusb-specific option (e.g. int hid_libusb_set_output_timeout(hid_device *dev, unsigned int timeout) to be used with hid_write and hid_send_feature_report would suffice.

P.S.: PR is welcome.

[elfmimi]

Thank you for your quick response. I'm afraid I'm not sure if I want to do a PR myself for now.
It's like I spent way too much effort into finding this out , and so I felt like leaving a short note here.

Easiest workaround is to use libhidapi-hidraw instead of libhidapi-libusb .

[Youw]

Easiest workaround is to use libhidapi-hidraw instead of libhidapi-libusb .

I'd say - a perfect one.

[Youw]

For a historical reference:
HIDRAW has this timeout hard-coded as well: https://github.com/torvalds/linux/blob/69b41ac87e4a664de78a395ff97166f0b2943210/drivers/hid/usbhid/hid-core.c#L931
And it is set to 5 seconds.

[mcuee]

I think we may want to promote the use of the hidraw backend for Linux and not the libusb backend.

[Youw]

Totally agree. But that does not exclude having a potential fix for libusb backend.

[oistra-ah]

Hi,
could please someone give a hint: is it possible to increase timeout for send_feature_report() on Windows? I also have a device, which takes a more than 5 sec sometimes for response. Thanks.
@Youw

[mcuee]

Hi, could please someone give a hint: is it possible to increase timeout for send_feature_report() on Windows? I also have a device, which takes a more than 5 sec sometimes for response. Thanks.

That is interesting, what is the device? The original issue only mentions Linux and only libusb backend.

[oistra-ah]

Hi, could please someone give a hint: is it possible to increase timeout for send_feature_report() on Windows? I also have a device, which takes a more than 5 sec sometimes for response. Thanks.

That is interesting, what is the device? The original issue only mentions Linux and only libusb backend.

Sorry for asking in this topic. I've just created #670. Device is a tiny USB climate sensor on the stm32f4