A few interface issue/Bug?

[KelvinChung2000]

Self checks

• I have searched existing issues, including closed issues.
• I have read the Contributing Guide.
• This is a bug report, not a question, feature request, or usage discussion.
• I will include logs, screenshots, or command output where they help reproduce the issue.
• I have removed secrets, private paths, tokens, and other sensitive information from the report.

ECOS Studio version

v0.1.0-alpha.8

Affected area

GUI

How did you get ECOS Studio?

Release AppImage

Where are you running ECOS Studio?

Native Linux desktop

Environment

PopOs 22.04, basiclly Ubuntu 22.04

ECC, PDK, and resource context

No response

Steps to reproduce

After running synthesis, the log is back up to the top, not showing the last message

Codex installed but is not detected

This is my codex path: /home/kelvin/.npm-global/bin/codex

But that being said, I much prefer Claude Code. But this sort of itergration is an minor issue. But I might actually prefer to not having it occpuying something by default. Since the use case inside the backend is minimal.

Yosys is not running correctly.

It is likely because it picked up my system yosys, not the app image yosys?

It would be nice to have the macro setting page added to the configuration.

In the configuration when I click save the on the configuration dashboard, the save does not happen.

And when I click the save button

It revet back.

Expected behavior

...

Actual behavior

...

Logs, screenshots, or recordings

No response

Additional context

No response

[KoEkko]

Thanks for the screenshots. This is several separate items on v0.1.0-alpha.8 — please try v0.1.0-alpha.9 first (Yosys install was broken on alpha.8, see #174).

1. Log jumps to the top after synthesis — real GUI bug. After the step finishes the viewer reloads the full log and does not stay at the tail. Workaround: expand the log and scroll down, or open the on-disk Synthesis_yosys log. We’ll fix the follow-tail behavior.

2. Codex not detected — expected for a desktop-launched AppImage: it does not inherit your shell PATH, so ~/.npm-global/bin/codex is invisible. Use 选择本地 codex, or:

   ECOS_AGENT_CODEX_BIN=/home/kelvin/.npm-global/bin/codex ./ECOS-Studio_*.AppImage
   Agent currently requires Codex. Claude Code support is planned; it is not in this release. Backend RTL-to-GDS does not need Agent, so you can ignore that panel for now.
   

3. Yosys / ABC crash — the log shows system Yosys (0.50+56 / /usr/local/bin/yosys-abc), not the managed runtime. After upgrading, install Yosys in Resource Manager and re-run synthesis.

4. Macro settings — not on the Configuration dashboard. They are on the Floorplan step config (macro_placer). We can consider surfacing a subset later.

5. Save reverts — alpha.8 GUI writes parameters.json, while ECC treats the TOML in home/ as source of truth, so refresh reloads the old top module. The proper fix is on main (feat(gui): prefer home/ecc.toml for workspace parameters #202) and is not in the alpha.9 AppImage yet. Until the next release, edit the TOML under workspace/home/ directly.

[KelvinChung2000]

For case 2, you might want to consider a few extra things. For example, I don't use Bash but Fish, so I guess even if I do a local install, it still won't be on my normal path (I guess). Since the install method and location are relatively fixed, you can hardcode and do a first lookup in those directories; then it would work across multiple methods.

As a UI suggestion, if the agent panel isn't used, you could use that space to show other information. Like even a larger log showing space is still better. Another good one is to add a terminal. This could be tied to a standard terminal or a REPL for ECC.

[KoEkko]

Both good suggestions — here's where we'll take them.

Codex lookup. One clarification first: this isn't fish-specific. A desktop-launched
AppImage doesn't inherit any interactive shell's PATH — the desktop session only reads
the login environment (~/.profile, ~/.config/environment.d/, …), never ~/.bashrc
or ~/.config/fish/. Bash users who add paths in .bashrc hit exactly the same wall;
your ~/.npm-global/bin entry lives in your fish config, where the app can't see it.

Your proposed fix is the right one, and it's what we'll implement. Today the resolution
order is: saved setting → ECOS_AGENT_CODEX_BIN → managed install → process PATH —
that last step is what fails in your screenshot. We'll insert a well-known-locations
probe before the PATH scan, covering the standard install methods:

• ~/.npm-global/bin — npm with a user prefix (your setup)
• ~/.nvm/versions/node/*/bin — nvm (glob, since the directory is versioned)
• ~/.volta/bin, ~/.bun/bin, ~/.local/share/pnpm
• ~/.local/bin, /usr/local/bin, /opt/homebrew/bin

That makes detection work across npm/nvm/volta/bun/pnpm/brew installs regardless of
shell or launcher — tracked in #220. Meanwhile, the Install button in the Agent
panel sidesteps PATH entirely (managed binary + saved setting), so it works on any
shell today.

Agent panel space. Agreed — we'll add a collapse so the panel's space goes back to
the step log when the Agent isn't in use (#222). A terminal or ECC REPL in that slot is
a good idea too, but it's feature-sized (PTY host in the main process, and an
interactive mode in the runtime for the REPL) — tracking it in #221.