PR #53 introduced a regression into boost pool

[ni-vzavalishin]

TL;DR: PR #53 seems to be based on a wrong assumption. In fact,simple_segregated_storage in its present form is fundamentally uncapable of handling mixed partition sizes, and cannot be "fixed" to allow that handling without a major change of the implementation. The linked-to Example 4.1, which was a starting point for the "fix", also looks simply wrong and so does the fix itself.

There is a new PR here, which contains further details, including

• the demonstration of the regression (the 1st commit of the new PR),
• the reverting of the code change (2nd commit of the new PR) and
• dropping of the two failing uint tests introduced by PR Fix the Broken malloc_n Behavior #53, which are believed to be incorrect.