Panic while reading xls

[prokie]

Hi When I try to open a workbook on one of my xls files I get the following error.

fn main() -> Result<()> {
    let mut workbook: Xls<_> = open_workbook("bla.xls")?;
    Ok(())
}

thread 'main' panicked at calamine-0.26.1/src/cfb.rs:362:9:
assertion `left == right` failed: i=3062, len=4334
  left: 0
 right: 3
stack backtrace:
   0:     0x7f7c0cfbab35 - <std::sys::backtrace::BacktraceLock::print::DisplayBacktrace as core::fmt::Display>::fmt::h358afad87e02ca76
   1:     0x7f7c0cfef77b - core::fmt::write::hb19b5b269a2fe458
   2:     0x7f7c0cfb8c1f - std::io::Write::write_fmt::he5a92676a45ef09d
   3:     0x7f7c0cfbbc81 - std::panicking::default_hook::{{closure}}::h3bff550b24d93725
   4:     0x7f7c0cfbb95c - std::panicking::default_hook::hd53b1b06d2b99687
   5:     0x7f7c0cfbc251 - std::panicking::rust_panic_with_hook::h9fdd87cddb2763da
   6:     0x7f7c0cfbc147 - std::panicking::begin_panic_handler::{{closure}}::h089783ab6b5cba45
   7:     0x7f7c0cfbaff9 - std::sys::backtrace::__rust_end_short_backtrace::hed34776d77ef7922
   8:     0x7f7c0cfbbdd4 - rust_begin_unwind
   9:     0x7f7c0ced69d3 - core::panicking::panic_fmt::h300583f35f37447a
  10:     0x7f7c0ced6dcf - core::panicking::assert_failed_inner::hafb0e3d63cb01ba6
  11:     0x7f7c0ced3dcf - core::panicking::assert_failed::h5549a7e67ae6daf2
  12:     0x7f7c0cf7f447 - calamine::cfb::decompress_stream::he271a08c0ffe52cb
  13:     0x7f7c0cf0afe7 - <alloc::vec::into_iter::IntoIter<T,A> as core::iter::traits::iterator::Iterator>::try_fold::hdf21cd55dade2c87
  14:     0x7f7c0ceedac1 - alloc::vec::in_place_collect::from_iter_in_place::h7d46edc2f8f2af80
  15:     0x7f7c0ceea3b9 - <alloc::collections::btree::map::BTreeMap<K,V> as core::iter::traits::collect::FromIterator<(K,V)>>::from_iter::hc1b6dd616abfaa91
  16:     0x7f7c0cf03bf2 - calamine::vba::VbaProject::from_cfb::h275adceebcd448df
  17:     0x7f7c0cee6b0d - calamine::xls::Xls<RS>::new_with_options::hed3b834fcadb6634
  18:     0x7f7c0cee63ec - calamine::open_workbook::hfc42529ea9511aa1
  19:     0x7f7c0cf0461f - pontus::main::hb4eed906d3c6d221
  20:     0x7f7c0cf10023 - std::sys::backtrace::__rust_begin_short_backtrace::h998b547d7489787b
  21:     0x7f7c0cefbc8d - std::rt::lang_start::{{closure}}::h2e2caad5b5a6f960
  22:     0x7f7c0cfb3af7 - std::rt::lang_start_internal::h93b3b742566fb30c
  23:     0x7f7c0cf06255 - main

Does this mean that the excel file is broken or why is it crashing?

[prokie]

It seems that the VBA script was the issue, I removed it and now the xls file parses without issue.

[sftse]

Can you provide a test case to investigate the issue?

[prokie]

Yes sure, I will see if I can remove everything sensitive in the xls and upload it here.

[prokie]

I am trying to remove all the confidential information, but is not easy. But the issue seems to be related to whitespace in VBA script.

If I add a newline in the VBA script the parsing will succeed every time, but if I add too many newlines I get the error at line 362 in cfb.rs consistently. There are two different failing behaviors, either it fails and left is always 2 or it fails and left is always 0.

[sftse]

If the issue is solely in the VBA script, it should be feasible to extract it using the rust-cfb crate into a separate CFB file. If after doing that calamine fails at the same line, this might help narrow down the cause.

[prokie]

Okay thanks, I will try that. I will also keep working on getting an example xls w/o confidential info..

[prokie]

@sftse Can you give example on how to do that using rust-cfb?

[sftse]

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let in_path = "foo.xls";
    let out_path = "bar.xls";
    let mut original = cfb::open(in_path)?;
    let version = original.version();
    let out_file = File::create(out_path).unwrap();
    let mut duplicate = cfb::CompoundFile::create_with_version(version, out_file)?;
    let mut stream_paths = Vec::<std::path::PathBuf>::new();
    for entry in original.walk() {
        if entry.path().to_str().unwrap().contains("VBA") {
            if entry.is_storage() {
                if !entry.is_root() {
                    duplicate.create_storage(entry.path())?;
                }
                duplicate.set_storage_clsid(entry.path(), entry.clsid().clone())?;
            } else {
                stream_paths.push(entry.path().to_path_buf());
            }
        }
    }
    for path in stream_paths.iter() {
        std::io::copy(
            &mut original.open_stream(path)?,
            &mut duplicate.create_new_stream(path)?,
        )?;
    }
    Ok(())
}

[prokie]

Thanks for that source code. I ran basically the same but added the xls read from calamine afterwards and it still gives the same error,

use std::fs::File;

use calamine::{open_workbook, Xls};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    env_logger::init();
    let in_path = "foo.xls";
    let out_path = "bar.xls";
    let mut original = cfb::open(in_path)?;
    let version = original.version();
    let out_file = File::create(out_path).unwrap();
    let mut duplicate = cfb::CompoundFile::create_with_version(version, out_file)?;
    let mut stream_paths = Vec::<std::path::PathBuf>::new();
    for entry in original.walk() {
        if entry.path().to_str().unwrap().contains("VBA") {
            if entry.is_storage() {
                if !entry.is_root() {
                    duplicate.create_storage(entry.path())?;
                }
                duplicate.set_storage_clsid(entry.path(), entry.clsid().clone())?;
            } else {
                stream_paths.push(entry.path().to_path_buf());
            }
        }
    }
    for path in stream_paths.iter() {
        std::io::copy(
            &mut original.open_stream(path)?,
            &mut duplicate.create_new_stream(path)?,
        )?;
    }
    let _: Xls<_> = open_workbook("bar.xls")?;

    Ok(())
}

assertion `left == right` failed: i=3062, len=4338
  left: 0
 right: 3

[sftse]

This reduced cfb file should have most of the sensitive parts removed. If you look through the remaining information and find it acceptable to publish, can you upload it as a test case?

[jmcnamara]

Without a reproducible test file the issue will have to be closed as "can't fix".

[jmcnamara]

Closing due to a lack of a reproducible file. Will reopen/reinvestigate if a file is provided.