read: segment too long (6182) with max (35872) for codeblock 0 (p=19, b=2, r=5, c=1)

[gcode-importer]

Originally reported on Google Code with ID 284

The attached file (see issue 135 for construction) fails to decode in current openjpeg:

read: segment too long (6182) with max (35872) for codeblock 0 (p=19, b=2, r=5, c=1)
[ERROR] Failed to decode.
[ERROR] Failed to decode tile 1/1
ERROR -> opj_decompress: failed to decode image!

Reported by malaterre on 2014-03-12 15:40:04

---

- _Attachment: [issue135.trans.j2k](https://storage.googleapis.com/google-code-attachments/openjpeg/issue-284/comment-0/issue135.trans.j2k)_

[gcode-importer]

I suspect issue comes from r2266. Indeed code read as:

                                if (l_current_data + l_seg->newlen > p_src_data + p_max_length)
{
                                        return OPJ_FALSE;
                                }
[...]
                                if ((l_cblk->data_current_size + l_seg->newlen) > l_cblk->data_max_size)
{
                                    OPJ_BYTE* new_cblk_data = (OPJ_BYTE*) opj_realloc(l_cblk->data,
l_cblk->data_current_size + l_seg->newlen);

Reported by malaterre on 2014-03-12 15:41:46

• Labels added: Milestone-Release2.1

[gcode-importer]

Reported by malaterre on 2014-03-14 14:04:28

• Status changed: Started

[tfmorris]

I have no idea if the above analysis is correct, but just to update the references:
r2262 == commit b65bf48

Code snippet above in diff: b65bf48#diff-1bfcd820328c9d73ee9052a650febf3eR1114

Code snippet in current master: https://github.com/uclouvain/openjpeg/blob/master/src/lib/openjp2/t2.c#L1191

[tfmorris]

I have additional images that demonstrate this problem if it would help with the debug. JPEG2000 is used extensively by the Internet Archive and OpenJPEG only appears to be able to decode about half the pages there. The kdu_expand program doesn't have an issue with them.

[tfmorris]

Oops! Ignore my comment. My problem was due to an problem further upstream in the processing pipeline which was feeding truncated JP2K data to OpenJPEG. Now opj_decompress seems to be working find as a drop-in replacement for kdu_expand.