[windows] Server not running on old CPUs

[afucher]

To upvote this issue, give it a thumbs up. See this list for the most upvoted issues.

Describe the bug

When running new releases on Windows (v0.39.0) it is failing due to CPU features required by the executable.
ECA stderrr:

Downloading eca from https://github.com/editor-code-assistant/eca/releases/download/0.114.0/eca-native-windows-amd64.zip to c:\Users\user\.vscode\extensions\editor-code-assistant.eca-0.39.0\eca-native-windows-amd64.zip
ECA artifact downloaded to c:\Users\user\.vscode\extensions\editor-code-assistant.eca-0.39.0\eca-native-windows-amd64.zip
[VSCODE] spawning server: c:\Users\user\.vscode\extensions\editor-code-assistant.eca-0.39.0\eca.exe with args: server --log-level debug
The current machine does not support all of the following CPU features that are required by the image: [CX8, CMOV, FXSR, MMX, SSE, SSE2, SSE3, SSSE3, SSE4_1, SSE4_2, POPCNT, LZCNT, AVX, AVX2, BMI1, BMI2, FMA, F16C].

Please rebuild the executable with an appropriate setting of the -march option.
[VSCODE] server process closed: code=1 signal=null

To Reproduce

Steps to reproduce the behavior:

1. Update ECA on windows
2. Try to run it

Expected behavior

That old CPUs are supported or have any information about the requirements to run ECA.

Doctor

$ wmic cpu get name
Name
Intel(R) Core(TM) i5-3570K CPU @ 3.40GHz

Additional context
I didn't have time to investigate more, but maybe is related to this commit 4aa0bc0 that changes the graalvm action

[afucher]

More additional info:
Seems that is possible to pass to the graalvm compiler the target CPU version
native-image -march=x86-64-v2
Mine is x86-64-v2 and based on the feature that is asking seems that is targeting x86-64-v3

[afucher]

I tried the new version but my Windows is not able to run the eca.exe locally.
I've built the native client locally and it works

/doctor

ECA version: 0.114.1
Server cmd: 
Workspaces: D:\Projetos\eca
Default model: openrouter/arcee-ai/trinity-large-preview:free (API key (sk-**********173))
Logged providers: 
openai: Subscription (oauth), expires 2026-03-09 15:22
Relevant env vars: 
Credential files: None found

Looking to differences between my local build and the pipeline, I saw that the pipeline does a UPX compression in the executable and it has a workaround to make it work (ref), but I found that this workaround seems to not be working anymore upx/upx#670 (comment)