Herb sometimes hangs when parsing an HTML file

[spect88]

Hi Marco!

When I try to parse the default 404 page in Rails, herb works most of the time, but sometimes it hangs (stuck at 100% CPU and I have to kill it)

To reproduce:

# Try this a couple more times if it exits properly
for i in {1..10}; do herb parse 404.html > /dev/null; done

I use herb (0.1.1-arm64-darwin) on ruby 3.4.3 (2025-04-14 revision d0b7e5b6a0) +PRISM [arm64-darwin23].

I've also seen an occasional segfault (when parsing the same file).

Another thing, which might or might not be related is that when parsing the same HTML file using the Ruby API, I sometimes get empty <style> tag:

    │   │   │   │   ├── @ HTMLElementNode (location: (14:4)-(96:12))
    │   │   │   │   │   ├── open_tag:
    │   │   │   │   │   │   └── @ HTMLOpenTagNode (location: (14:4)-(14:11))
    │   │   │   │   │   │       ├── tag_opening: "<" (location: (14:4)-(14:5))
    │   │   │   │   │   │       ├── tag_name: "style" (location: (14:5)-(14:10))
    │   │   │   │   │   │       ├── tag_closing: ">" (location: (14:10)-(14:11))
    │   │   │   │   │   │       ├── children: []
    │   │   │   │   │   │       └── is_void: false
    │   │   │   │   │   │
    │   │   │   │   │   ├── tag_name: "style" (location: (14:5)-(14:10))
    │   │   │   │   │   ├── body: (1 item)
    │   │   │   │   │   │   └── @ HTMLTextNode (location: (14:11)-(96:4))
    │   │   │   │   │   │       └── content: ""
    │   │   │   │   │   │
    │   │   │   │   │   ├── close_tag:
    │   │   │   │   │   │   └── @ HTMLCloseTagNode (location: (96:4)-(96:12))
    │   │   │   │   │   │       ├── tag_opening: "</" (location: (96:4)-(96:6))
    │   │   │   │   │   │       ├── tag_name: "style" (location: (96:6)-(96:11))
    │   │   │   │   │   │       └── tag_closing: ">" (location: (96:11)-(96:12))
    │   │   │   │   │   │
    │   │   │   │   │   └── is_void: false

[marcoroth]

Thanks for the report!

I'm not a 100% sure what causes this, but I'm pretty sure it has to do with memory leaks in the Ruby C-Extension itself.

I'll have to do some more tracing to make sure and confirm that's what's going on. But that would also explain why it's random and not 100% reproducible on every run.

/cc @dennispaagman as I'm pretty sure that's a very similar bug to what you reported to me via email!

[dennispaagman]

Yeah sounds very similar! Wondering if it has to do with the style tags, as the files that crash for me also have inline styles.

[marcoroth]

Small update, I think I was able to scope down the issue a bit here. It has to do with the buffer Herb uses. The issue seems to occur whenever the content exceeds the buffer capacity and it tries to increase its capacity.

Currently the buffer sets its initial capacity to 1024 bytes and doubles it every time it needs to hold more than it's currently able to.

Using the 404 page in Rails you can somewhat reliably reproduce it, which is great. Thanks for that @spect88!

You can increase the success rate of it crashing when you set down the initial buffer capacity to 512 bytes. It also works the other way: I wasn't able to crash it once when I increased the initial buffer capacity to 8192 bytes.

herb/src/buffer.c

Line 11 in 623569d

buffer->capacity = 1024;

So it definitely has something to do with, when text content, attribute values (or similar) exceed the initial 1024 buffer capacity and it trying to reallocate memory to increase the buffer capacity. I'll keep digging.

[marcoroth]

I think I was able to resolve this using #122 and I'm going to release v0.2.0 with this.

I'm going to keep this open to make sure it also works in your case.

[marcoroth]

Hey @spect88 and @dennispaagman, if you get a chance, I'd love to hear if you still see the same behavior with v0.2.0. Thank you!

[spect88]

I can't reproduce the issue anymore 👍
Thanks for the quick fix!