Re: [Openocd-development] Change in openocd[master]: jlink libusb1 driver with libusb1 common driver interface.
I think that Xiofan suggest the good direction of developement. So I think to proceed in this way: - Add to the configuration step the possibility to choose which usb library to use. - Create an abstraction layer to the usb library with all the necessary functions for the drivers. - Modify the drivers to use the abstraction layer instead directly the library. Do you think it's reasonable? Mauro Gamba On 20 October 2011 07:02, Xiaofan Chen xiaof...@gmail.com wrote: On Wed, Oct 19, 2011 at 11:55 PM, Spencer Oliver s...@spen-soft.co.uk wrote: On 19 October 2011 16:38, Mauro Gamba maurill...@gmail.com wrote: Sorry for patch errors. I started to patch the jlink driver to use libusb-1 because libusb-0 is not developed further. I haven't done speed tests until now. http://openocd.zylin.com/33 adds libusb-1.0 support to the jlink. Just wondering if anyone had any input on this ? I believe the approach to only uss libusb-1.0 for J-Link is not a good approach. My idea is to have both options, just like urjtag. When libusb-1.0 is available and specified by the user, it should use libusb-1.0, other wise, it will fall back to libusb-0.1. Benefits of providing both: 1) Make the regression testing easier. 2) Make J-Link to work on platforms where libusb-1.0 is not available, like Solaris/NetBSD/OpenBSD, and older version of FreeBSD, and maybe some embedded Linux platform, and Windows 2000. 3) Make users who prefer to use libusb-0.1 can use libusb-0.1, say Windows users who prefer to keep both Segger driver (to use IAR/Keil/etc) and OpenOCD -- they can use libusb-win32 filter driver. Take note libusb-1.0 Windows does not support libusb0.sys backend now and that makes it not working with the filter driver. -- Xiaofan ___ Openocd-development mailing list Openocd-development@lists.berlios.de https://lists.berlios.de/mailman/listinfo/openocd-development
Re: [Openocd-development] Change in openocd[master]: jlink libusb1 driver with libusb1 common driver interface.
Mauro Gamba wrote: I think to proceed in this way: - Add to the configuration step the possibility to choose which usb library to use. Use pkg-config to check for all libraries that can work, and use with the first one that is found. Check for libusb-1.0 first. You will need to study more autoconf than you needed before. You can of course also ask on the list if you are uncertain about something. - Create an abstraction layer to the usb library with all the necessary functions for the drivers. Yes and no. all the neccessary functions should not be created at once. Start with an abstraction layer that works for one driver but still allows upward compatibility with the other drivers. This may be more difficult than you think. You will have to analyze each driver more in depth, and I guess it will take a bit of time. The above two items shall be in the first commit. Please do not wait for the complete driver analysis before you make the commit, please start with creating an abstraction layer that is sufficient for only one driver, push that to Gerrit, and see what the feedback is. This is OK to do even if you want to do more work on the commit, but it is important to get review feedback early, so that you do not have to do double work. Of course, remember to pay close attention to simple details like checking all possible errors and code style. - Modify the drivers to use the abstraction layer instead directly the library. Do you think it's reasonable? Yes, but it is more work and more complicated, so it is even more important than before to only start small and then gather feedback. Thanks again for working on this! I'm really looking forward to your commits! //Peter ___ Openocd-development mailing list Openocd-development@lists.berlios.de https://lists.berlios.de/mailman/listinfo/openocd-development
Re: [Openocd-development] Change in openocd[master]: jlink libusb1 driver with libusb1 common driver interface.
Spencer Oliver wrote: http://openocd.zylin.com/33 adds libusb-1.0 support to the jlink. Just wondering if anyone had any input on this ? Yes, plenty. //Peter ___ Openocd-development mailing list Openocd-development@lists.berlios.de https://lists.berlios.de/mailman/listinfo/openocd-development
Re: [Openocd-development] Change in openocd[master]: jlink libusb1 driver with libusb1 common driver interface.
On Wed, Oct 19, 2011 at 11:55 PM, Spencer Oliver s...@spen-soft.co.uk wrote: On 19 October 2011 16:38, Mauro Gamba maurill...@gmail.com wrote: Sorry for patch errors. I started to patch the jlink driver to use libusb-1 because libusb-0 is not developed further. I haven't done speed tests until now. http://openocd.zylin.com/33 adds libusb-1.0 support to the jlink. Just wondering if anyone had any input on this ? I believe the approach to only uss libusb-1.0 for J-Link is not a good approach. My idea is to have both options, just like urjtag. When libusb-1.0 is available and specified by the user, it should use libusb-1.0, other wise, it will fall back to libusb-0.1. Benefits of providing both: 1) Make the regression testing easier. 2) Make J-Link to work on platforms where libusb-1.0 is not available, like Solaris/NetBSD/OpenBSD, and older version of FreeBSD, and maybe some embedded Linux platform, and Windows 2000. 3) Make users who prefer to use libusb-0.1 can use libusb-0.1, say Windows users who prefer to keep both Segger driver (to use IAR/Keil/etc) and OpenOCD -- they can use libusb-win32 filter driver. Take note libusb-1.0 Windows does not support libusb0.sys backend now and that makes it not working with the filter driver. -- Xiaofan ___ Openocd-development mailing list Openocd-development@lists.berlios.de https://lists.berlios.de/mailman/listinfo/openocd-development
Re: [Openocd-development] Change in openocd[master]: jlink libusb1 driver with libusb1 common driver interface.
On Thu, Oct 20, 2011 at 1:02 PM, Xiaofan Chen xiaof...@gmail.com wrote: On Wed, Oct 19, 2011 at 11:55 PM, Spencer Oliver s...@spen-soft.co.uk wrote: http://openocd.zylin.com/33 adds libusb-1.0 support to the jlink. Just wondering if anyone had any input on this ? I believe the approach to only uss libusb-1.0 for J-Link is not a good approach. My idea is to have both options, just like urjtag. When libusb-1.0 is available and specified by the user, it should use libusb-1.0, other wise, it will fall back to libusb-0.1. Benefits of providing both: 1) Make the regression testing easier. 2) Make J-Link to work on platforms where libusb-1.0 is not available, like Solaris/NetBSD/OpenBSD, and older version of FreeBSD, and maybe some embedded Linux platform, and Windows 2000. 3) Make users who prefer to use libusb-0.1 can use libusb-0.1, say Windows users who prefer to keep both Segger driver (to use IAR/Keil/etc) and OpenOCD -- they can use libusb-win32 filter driver. Take note libusb-1.0 Windows does not support libusb0.sys backend now and that makes it not working with the filter driver. Again, I will point the link to urjtag. http://urjtag.git.sourceforge.net/git/gitweb.cgi?p=urjtag/urjtag;a=tree;f=urjtag/src/tap/usbconn;hb=HEAD In terms of the simple wrapper of libusb-1.0, I think this is a good first try. On the other hand, the function jtag_libusb_match() can probably take one more parameter serial_number so that some debuggers can benefits from the support of multiple units. -- Xiaofan ___ Openocd-development mailing list Openocd-development@lists.berlios.de https://lists.berlios.de/mailman/listinfo/openocd-development
