Re: [PATCH v5 11/17] tests/docker: strip remaining caching logic from docker.py

["Alex Bennée" <alex.bennee@linaro.org>]

Pierrick Bouvier <pierrick.bouvier@oss.qualcomm.com> writes:

> On 9/17/2026 3:18 PM, Richard Henderson wrote:
>> On 9/17/26 12:00, Pierrick Bouvier wrote:
>>> On 9/17/2026 2:49 PM, Alex Bennée wrote:
>>>> how much time? it flies past when I run a rebuild.
>>>>
>>>
>>> For a no-op tcg-tests build (i.e. with tests already built):
>>>
>>> - With current patch:
>>> time ninja -C tcg-tests
>>> ...
>>> verbose pull and build
>>> 320 lines of output
>>> ...
>>> real    0m15.817s
>>> user    0m1.342s
>>> sys     0m0.840s

  ✗  time ninja
  [1/114] Generating qemu-version.h with a custom command (wrapped by meson to capture output)

  ________________________________________________________
  Executed in  515.16 millis    fish           external
     usr time  433.82 millis  968.00 micros  432.85 millis
     sys time   86.05 millis  253.00 micros   85.80 millis

Not sure whats going on there then. The whole tcg suite completes in
under a minute (and would be shorter if i386-linux-user-test-i386-fprem
and riscv64-softmmu-plugin-discons-with-interruptedmemory where in the
thorough test suite).

>> 
>> The variance from the middle of the pacific ocean, and associated
>> network latency, is between 20 and 27 seconds.
>> 
>> Amusingly, with the network actively disconnected (airplane mode), the
>> no-op build completes in 1 to 2 seconds.
>>
>
> Funny, on my side if hangs a long time (timeout 30s), then retry 3 times
> per image. I tried it on WSL on my windows laptop.
>
>> 
>> r~

-- 
Alex Bennée
Virtualisation Tech Lead @ Linaro