When writing compound file, the result is sometimes incorrect

*Imported Issue originally submitted by a former Bitbucket user*

I am using the pole library to write and read outlook MSG files (to convert to/from mime). Until today it was good. But today I found the following: one of the attachment data stream has invalid "cluster" chain. When writing the file, all looks OK in debugger. But the resulting file is incorrect: far manager plugin refuses to open it; when I run my program against that file to convert to mime, it fails also: the problematic stream is attach_version1.0_#00000000/substg1.0_37010102 Its "correct" size is 24931929 bytes. When I read this stream into buffer, the byte count is only 11057664.

Additional info: streamIO::blocks.size() is 48696 when writing the stream. After the read it is 21597. I believe it should be the same 48696. That is why I think that the stream has invalid chain.

Please help me to debug the issue. I tried myself but it is too contrived.

*Note: This is a split zip archive. The `.z01.zip` extension has to be renamed to `.z01`, and `.z02.zip` to `.z02`. GitHub wouldn't otherwise accept it.*

[email8.zip](https://github.com/KubaO/pole/files/4692000/email8.zip)

[email8.z02.zip](https://github.com/KubaO/pole/files/4691997/email8.z02.zip)

[email8.z01.zip](https://github.com/KubaO/pole/files/4691995/email8.z01.zip)

Start by reproducing the failure with the attached split archive after renaming the .z01.zip and .z02.zip extensions as described. Compare the attach_version1.0_#00000000/substg1.0_37010102 stream's expected 24931929-byte size and streamIO::blocks.size() value of 48696 with the values after reading; done means the resulting file opens and the stream reads completely.

Written by the indexing model from the issue text.