pylink-square==1.5.0 fails to open J-Link by serial on Ubuntu 22.04 (JLink V8.74); 1.4.0 works

[YoannD42]

Hello,

After upgrading from pylink-square==1.4.0 to 1.5.0, calling JLink.open(serial_no=...) over USB consistently fails with pylink.errors.JLinkException: No emulator with serial number … found, even though the J-Link is visible via lsusb and works with JLinkExe. Downgrading back to 1.4.0 restores expected behavior. Nothing else (Docker run, udev rules, permissions) changed between images beyond the pylink version bump.

Environment
• OS: Ubuntu 22.04 (inside Docker)
• Python: 3.10
• pylink-square:
• 1.4.0 → works
• 1.5.0 → fails (regression)
• J-Link SDK: V8.74 (installed under /opt/SEGGER/JLink_V874/libjlinkarm.so.8)
• Hardware (examples from lsusb):
1366:0101 SEGGER J-Link PLUS
1366:1061 SEGGER J-Link
• Example device S/Ns observed: 821008920, 821008926
• JLinkExe output (works):

SEGGER J-Link Commander V8.74 Connecting to J-Link via USB...O.K. S/N: 821008926 ... Device "EFM32GG11B820F2048" selected. Connecting to target via SWD ... OK

1. Install SEGGER J-Link V8.74 (Linux), so the library is at /opt/SEGGER/JLink_V874/libjlinkarm.so.8.
2. pip install pylink-square==1.5.0
3. Run:

import pylink j = pylink.JLink() j.open(serial_no=821008920) # any connected J-Link serial

4.  Observe failure (exception below).
   

pylink.errors.JLinkException: No emulator with serial number 821008920 found.

5. Downgrade to pylink-square==1.4.0 and re-run the same snippet → succeeds.

Notes / Possibly Relevant Change

This regression coincides with the 1.5.0 @bbrown1867 change that dynamically decides whether to load a temporary copy of the J-Link DLL (use_tmpcpy) based on a parsed DLL version. I didn’t modify that setting explicitly; I only observe that 1.5.0 fails and 1.4.0 works under the same environment.

Happy to provide any extra diagnostics you’d like.

[hkpeprah]

Sorry, I forgot to respond to this. If you pass use_tmpcpy=True then it works right?

[YoannD42]

@hkpeprah My turn to apologise for the delay 🙏 , this is working if we use use_tmpcpy=True. We indeed can upgrade pylink, by setting that.

Any hint ? Not quite sure what would be the impact of using that flag, versus not to.

[hkpeprah]

Originally, the shared C library file was internally copied by the Python library to workaround a global lock that prevented using two debuggers at the same time. This was patched in a later revision of the shared C library, which removed the need for the temporary copy. The temporary copy also prevents working with custom XML files defining new chips or interfaces for flashing, so we defaulted the option to False.

Changing it to True will require you to copy over any custom XML files you have for custom devices. If you don't have any custom devices, then there should be no impact to your workflow.