Johan Corveleyn <jcor...@gmail.com> writes: > I'm not sure what the reason could be. Perhaps the dumpfile itself is > huge, or has something special ... Or maybe the memory footprint is > normal, but the machine you're running it on is low on available > memory.
Apparently, our svn_stream_readline() implementation may cause excessive memory allocations if the input contains nul bytes. While I can't say for sure, this could be the likely cause of the reported problem. For example, the dump file could have been resaved in UTF-16, or something like that. I committed r1866950 that should fix this specific problem: https://svn.apache.org/r1866950 The other possible, although less likely reason for this problem is that the dump file itself is a large binary file without any \n characters. I committed r1866951 that should make the parser more resilient to such files: https://svn.apache.org/r1866951 Full solution for the latter part, I think, would require introducing and implementing something like svn_stream_readline2() with a `limit` parameter, but this change should at least ensure that the parser doesn't choke in the described case with a binary file and other similar situations. I will try to nominate both of these fixes to our stable branches, once I have some time for that. Thanks, Evgeny Kotkov