term: clamp zero window size from Docker PTY#959
Merged
pkova merged 2 commits intoFeb 17, 2026
Conversation
When Docker allocates a PTY (tty: true + stdin_open: true) but nobody is attached, TIOCGWINSZ succeeds but reports 0x0. This causes unsigned underflow in row_l - 1 calculations throughout term.c and a decrement-underflow crash in Arvo's drum when it receives %blew [0 0]. Clamp to 80x24 defaults when ioctl returns zero dimensions, matching the existing fallback when ioctl fails. Also fix the initial default row_l from 0 to 24 to match col_l. Resolves urbit#159.
Fang-
previously approved these changes
Feb 9, 2026
Fang-
left a comment
Member
There was a problem hiding this comment.
Regardless of where the dimensions come from, zeroes in %blew are bogus, so clamping like this is good. Arguably 0 == col_l instead of !col_l would be semantically clearer, but it's C, so readers know.
Collaborator
|
What version of vere are you running when you blow up with your reproduction commands? We've been unable to reproduce on linux-x86_64 and linux-aarch64. |
Contributor
Author
The issue was observed with FWIW, for a long time I've been operating with this workaround. I tested the fix in this PR by cross-compiling from this branch for aarch64-linux-musl and booting a comet in Docker (via OrbStack on macOS) with Worth mentioning that the reproduction is timing-sensitive: once anyone |
Contributor
Author
Contributor
Author
|
Tested on a live x86_64 VPS (Ubuntu, Docker 28.1.1) — the same environment where the issue was originally observed. Steps:
# 1. Cross-compile
zig build -Dtarget=x86_64-linux-musl -Doptimize=ReleaseFast
# 2. Build Docker image
cp zig-out/x86_64-linux-musl/urbit docker/
docker build --platform linux/amd64 -t vere-test-x86 docker/
# 3. Transfer to VPS
docker save vere-test-x86 | gzip > /tmp/vere-test-x86.tar.gz
scp /tmp/vere-test-x86.tar.gz user@vps:/tmp/
# on VPS:
docker load < /tmp/vere-test-x86.tar.gz
# 4. Add to docker-compose.yml under services:
# test-comet:
# image: vere-test-x86
# volumes:
# - ~/urbit/test-comet:/urbit
# stdin_open: true
# tty: true
mkdir -p ~/urbit/test-comet && touch ~/urbit/test-comet/test.comet
# 5. Start without attaching
docker compose up -d test-cometResult: Comet booted successfully and has been running for 8+ hours with no crash or attach. Dojo prompt is active, gall apps loaded normally. |
pkova
added a commit
that referenced
this pull request
Feb 18, 2026
Collaborator
|
FYI I reverted this because of a regression in terminal behavior on at least macos-aarch64. look at all the strange empty linesthis is normalThis needs to be fixed if we want this to go in. |
Contributor
Author
Thanks, will investigate |
tomholford
added a commit
to tomholford/vere
that referenced
this pull request
Feb 18, 2026
When Docker allocates a PTY (tty: true + stdin_open: true) but nobody is attached, TIOCGWINSZ succeeds but reports 0x0. This causes a decrement-underflow crash in drum when it receives %blew [0 0]. Clamp to 80x24 defaults when ioctl returns zero dimensions, matching the existing fallback when ioctl fails entirely. Unlike the prior attempt (urbit#959), this does not change the initial row_l from 0 — that value serves as a sentinel throughout term.c to skip cursor positioning during early boot (checked by hija/loja and the spinner). Changing it to 24 caused blank lines in terminal output on macOS. Resolves urbit#159.
pkova
added a commit
that referenced
this pull request
Mar 13, 2026
## Summary - When Docker allocates a PTY (`tty: true` + `stdin_open: true`) but nobody is attached, `TIOCGWINSZ` succeeds but reports 0x0 - This causes a `decrement-underflow` crash in drum when it receives `%blew [0 0]` - Clamp to 80x24 defaults when ioctl returns zero dimensions (matching existing fallback when ioctl fails) This is a v2 of #959, which was merged then reverted due to a regression. The prior version also changed the initial `row_l` from 0 to 24, which broke terminal output on macOS — `row_l = 0` is a sentinel used by `hija`/`loja` and the spinner to skip cursor positioning during early boot. Setting it to 24 caused every `u3l_log()` call before the first `%blew` event to reposition the cursor, producing blank lines. This version keeps `row_l = 0` and only adds the clamping in `u3_term_get_blew`. Resolves #159. See also [urbit/urbit#4750](urbit/urbit#4750) ## Test results **Docker on x86_64 Linux VPS** (Ubuntu, `tty: true` + `stdin_open: true`, no attach): - Cross-compiled for `x86_64-linux-musl`, built Docker image, transferred to VPS - Fresh comet booted successfully, ran 8+ minutes with no crash - Container config matches the exact scenario that triggers the bug **macOS aarch64** (native build, PTY via `script`): - Fresh comet booted, full bootstrap completed - No spurious blank lines in boot output (the regression from #959 is fixed) - Early boot log lines use plain text (no cursor positioning), confirming `row_l = 0` sentinel works correctly
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
tty: true+stdin_open: true) but nobody is attached,TIOCGWINSZsucceeds but reports 0x0row_l - 1calculations throughout term.c and adecrement-underflowcrash in drum when it receives%blew [0 0]row_lfrom 0 to 24Resolves #159.
See also urbit/urbit#4750
Disclaimer
I made this change using a coding agent. Please pardon any slop.
Test plan
zig buildzig build -Dtarget=aarch64-linux-musl -Doptimize=ReleaseFasttty: true+stdin_open: trueand nodocker attach— ship boots without crashingdocker attach/ detach still works after bootReproduction steps