==============================================================
Guild: wafer.space Community
Channel: 📐 - Designing / 📝-project-template / `librelane-magicdrc` crashes
After: 2025-10-31 11:59 p.m.
Before: 2025-12-01 12:00 a.m.
==============================================================

[2025-11-19 7:20 a.m.] mole99



[2025-11-19 7:20 a.m.] mole99
No, that is not expected. `librelane-magicdrc` only skips the `KLayout.DRC` step. Streamout, however, should still be done with both magic and KLayout in order to compare the GDS.
Can you verify that both `56-magic-streamout` and `57-klayout-streamout` contain a GDS?


[2025-11-19 10:46 a.m.] _luke_w_
Haha ok, I'll dig in a bit more this evening! Thanks

{Reactions}
👌

[2025-11-19 7:09 p.m.] _luke_w_
so yeah, the `.magic.drc` has a size of 0 bytes, and the previous step failed, citing:

```
Cell nmos_1p2_02_R270_512x8m81 read from path /home/luke/proj/riscboy-180/gf180mcu/gf180mcuD/libs.ref/gf180mcu_fd_ip_sram/mag
File /home/luke/proj/riscboy-180/gf180mcu/gf180mcuD/libs.ref/gf180mcu_fd_ip_sram/mag/M1_PSUB_05_512x8m81.mag couldn't be read
Failure to read entire subtree of the cell.
Failed on cell M1_PSUB_05_512x8m81.
I/O error in writing file /home/luke/proj/riscboy-180/librelane/runs/RUN_2025-11-19_18-47-13/56-magic-streamout/chip_top.magic.gds.
File may be incompletely written.
[INFO] GDS Write Complete
```

That SRAM filename does look wrong to me. This similar filename does exist:

```
./gf180mcu/gf180mcuD/libs.ref/gf180mcu_fd_ip_sram/mag/M1_PSUB_05_512x8m81.mag
```

Not totally sure where the incorrect filename is coming from. I don't see it in my `config.yaml`


[2025-11-19 7:11 p.m.] _luke_w_
oh wait no, that filename (from `find`) does match the one that magic couldn't read 🤔


[2025-11-19 7:12 p.m.] mole99
Hm, maybe too many file handlers open? On what platform are you building?


[2025-11-19 7:12 p.m.] _luke_w_
ahhh that could be it


[2025-11-19 7:12 p.m.] _luke_w_
I'm on Ubuntu


[2025-11-19 7:13 p.m.] mole99
On Debian I haven't had this issue before 🤔


[2025-11-19 7:14 p.m.] _luke_w_
```
luke@triton:~/proj/riscboy-180/hdl$ cat /proc/sys/fs/file-max
9223372036854775807
luke@triton:~/proj/riscboy-180/hdl$ ulimit -Hn
1048576
luke@triton:~/proj/riscboy-180/hdl$ ulimit -Sn
1024
luke@triton:~/proj/riscboy-180/hdl$ 
```
wat


[2025-11-19 7:16 p.m.] _luke_w_
thanks that was a good tip, I'll try with an increased limit. Somehow never hit this before 😅


[2025-11-19 7:17 p.m.] mole99
I only thought of it because ReJ has hit it on macOS recently ^^

{Reactions}
👍

[2025-11-19 7:18 p.m.] _luke_w_
you do tend to develop these spidey senses over time on a project

{Reactions}
🕷️

[2025-11-19 7:40 p.m.] _luke_w_
It seems much happier now. Thank you very much for your help


[2025-11-19 7:54 p.m.] mole99
Glad I could help!


[2025-11-19 8:51 p.m.] mithro_
@Leo Moser (mole99) - Maybe LibreLane should have a check that ulimit is a reasonable value?


[2025-11-20 9:47 a.m.] mole99
Working on it: https://github.com/librelane/librelane/issues/818

{Embed}
https://github.com/librelane/librelane/issues/818
Check or set the maximum number of open file descriptors · Issue #...
Description Leading up to the first wafer.space shuttle, two users faced the issue that the max open file descriptors count was set too low by default. One user is on Ubuntu and the other on macOS,...
2025-11_media/818-D8A72


==============================================================
Exported 17 message(s)
==============================================================