There is an unconditional second call to Seek() after a possible call to ReallocLastSegment(). Looks like the second seek is only needed if realloc was done?
Created attachment 572403 [details] [diff] [review]
Looks like base pointer changed only on positive return from ReallocLastSegment(). Maybe we should add this to the doc for ReallocLastSegment() as well...?
Please could you tweak your hgrc to automatically add author info (guide here: http://blog.bonardo.net/2010/06/22/so-youre-about-to-use-checkin-needed) + add a commit message when attaching patches, since it makes pushing half a dozen checkin-neededs a lot easier. Thanks :-)
Created attachment 572589 [details] [diff] [review]
Proposed fix w/ proper metainfo
Sorry - added here...