[Bug] Hitting dimension arrays of non-standard size, 16 instead of 14

[sftse]

[2025-10-23T09:42:18Z DEBUG calamine::xls] parse_dimensions: [0, 0, 0, 0, 99, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0]
[2025-10-23T09:42:18Z DEBUG calamine::xls] parse_dimensions: [0, 0, 0, 0, 221, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0]
[2025-10-23T09:42:18Z DEBUG calamine::xls] parse_dimensions: [0, 0, 0, 0, 99, 1, 0, 0, 0, 0, 5, 0, 0, 0, 0, 0]
[2025-10-23T09:42:18Z DEBUG calamine::xls] parse_dimensions: [0, 0, 0, 0, 221, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0]
[2025-10-23T09:42:18Z DEBUG calamine::xls] parse_dimensions: [0, 0, 0, 0, 99, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0]
[2025-10-23T09:42:18Z DEBUG calamine::xls] parse_dimensions: [0, 0, 0, 0, 125, 2, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0]
[2025-10-23T09:42:18Z DEBUG calamine::xls] parse_dimensions: [0, 0, 0, 0, 221, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0]
[2025-10-23T09:42:18Z DEBUG calamine::xls] parse_dimensions: [0, 0, 0, 0, 123, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0]

Similar to #572, wrong worksheet dimensions should not be much of an obstacle to reading the rest of the information from the sheet.

Could make it on-demand just like reading the VBA.

[jmcnamara]

Thanks. Do you have a test file that demonstrates this issue?

[sftse]

Unfortunately not, client data.
Comparing to other dimension arrays from files that do not have issues, the numbers seem to line up and have sensible values, just the two extra 0s are out of place.

[jmcnamara]

Unfortunately not, client data.

If you tell me which of the field(s) are incorrect I can create a test file. Or you could maybe dbg!() the errant DIMENSIONS record without printing any other sensitive data. That would be enough to create a test file.

[sftse]

If we strip the last two 0s from the arrays, we get 14 bytes that match spec-conformant Dimensions.

The first three examples parsed that way are then

Dimensions { start: (0, 0), end: (354, 0) }
Dimensions { start: (0, 0), end: (476, 0) }
Dimensions { start: (0, 0), end: (354, 4) }

Seems like reasonable values. Here are some from other files from the test set

Dimensions { start: (0, 0), end: (54, 6) }
Dimensions { start: (0, 0), end: (0, 0) }
Dimensions { start: (0, 1), end: (61, 27) }
Dimensions { start: (0, 0), end: (191, 74) }
Dimensions { start: (2, 0), end: (264, 9) }

[jmcnamara]

If we strip the last two 0s from the arrays, we get 14 bytes that match spec-conformant Dimensions.

Ok. Got it. This is definitely outside the spec for any Biff version of Excel.

And is the length field for these DIMENSIONS records 0x0010 = 16? Also, is the file readible by Excel?

[jmcnamara]

Here is a test file:

gh577.xls

@sftse Could you check if this is the same as your issue.

This isn't readible by Excel.

And an example:

use calamine::{open_workbook, Error, Xls};

fn main() -> Result<(), Error> {
    let path = "tests/gh577.xls";

    let _excel: Xls<_> = open_workbook(path).expect(path);

    Ok(())
}

And the error message:

$ cargo run --example gh577
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.27s
     Running `target/debug/examples/gh577`

thread 'main' panicked at examples/gh577.rs:6:46:
tests/gh577.xls: Len { expected: 14, found: 16, typ: "dimensions" }
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

[sftse]

This hits the same error, but if Excel cannot read it will further investigate, maybe the root cause is something else.

[jmcnamara]

but if Excel cannot read it will further investigate,

Just to note that Excel may not be able to read my test file because I used a program that generates xls files and I modified the code to create a DIMENSIONS field with 4 bytes of reserved data instead of the specified 2. However, this may have affected the worksheet offset calculations within the file. And that may be the reason that Excel cannot read it. Is Excel able to read your 16 byte DIMENSIONS file?

[sftse]

Is Excel able to read your 16 byte DIMENSIONS file?

Have not checked but that was what I was going to look into.

[sftse]

Excel is able to read the file that triggered this issue without warnings. If I open the test file gh577.xls in the online version of excel it opens with a warning that it had to be repaired.

Further of note, libreoffice opens gh577.xls without any warning.

[jmcnamara]

So, overall gh577.xls should be acceptable as a test input for this issue, is that correct?