On Wed, Feb 26, 2014 at 7:13 PM, Philip Martin <phi...@codematters.co.uk>wrote:
> $ svnadmin create repo3 --fs-type fsx > $ m -f repo3/db/format > $ printf "1\nlayout sharded 2\n" > repo3/db/format > $ svn mkdir -mm file://`pwd`/repo3/A > Thanks for the reproduction recipe, Philip. > $ valgrind -q subversion/svnadmin/.libs/lt-svnadmin pack repo3 > ==13523== Syscall param write(buf) points to uninitialised byte(s) > ==13523== at 0x4A97110: __write_nocancel (syscall-template.S:82) > ==13523== by 0x4A6AAF3: apr_file_flush_locked (readwrite.c:317) > ==13523== by 0x4A6B10F: setptr (seek.c:25) > ==13523== by 0x4A6B2E5: apr_file_seek (seek.c:70) > ==13523== by 0x4111A0F: svn_io_file_seek (io.c:3553) > ==13523== by 0x50CB6FA: write_changes_container (pack.c:1490) > ==13523== by 0x50CBD28: write_changes_containers (pack.c:1600) > ==13523== by 0x50CC904: pack_range (pack.c:1862) > ==13523== by 0x50CD246: pack_log_addressed (pack.c:2040) > ==13523== by 0x50CD6A3: pack_rev_shard (pack.c:2149) > ==13523== by 0x50CD892: pack_shard (pack.c:2204) > ==13523== by 0x50CDE78: pack_body (pack.c:2333) > ==13523== Address 0x69a5a98 is 56 bytes inside a block of size 4,096 > alloc'd > ==13523== at 0x402C1F0: malloc (vg_replace_malloc.c:291) > ==13523== by 0x4A6DEEB: pool_alloc (apr_pools.c:1463) > ==13523== by 0x4A6E067: apr_palloc_debug (apr_pools.c:1504) > ==13523== by 0x4A69573: apr_file_open (open.c:211) > ==13523== by 0x410BC09: file_open (io.c:351) > ==13523== by 0x411163D: svn_io_file_open (io.c:3428) > ==13523== by 0x50C8226: initialize_pack_context (pack.c:261) > ==13523== by 0x50CCFDC: pack_log_addressed (pack.c:1996) > ==13523== by 0x50CD6A3: pack_rev_shard (pack.c:2149) > ==13523== by 0x50CD892: pack_shard (pack.c:2204) > ==13523== by 0x50CDE78: pack_body (pack.c:2333) > ==13523== by 0x50D9368: with_some_lock_file (transaction.c:250) > ==13523== > ==13523== Syscall param write(buf) points to uninitialised byte(s) > ==13523== at 0x4A97110: __write_nocancel (syscall-template.S:82) > ==13523== by 0x4A6AAF3: apr_file_flush_locked (readwrite.c:317) > ==13523== by 0x4A6ABB3: apr_file_flush (readwrite.c:340) > ==13523== by 0x4A69252: apr_unix_file_cleanup (open.c:77) > ==13523== by 0x4A6F384: apr_pool_cleanup_run (apr_pools.c:2343) > ==13523== by 0x4A69698: apr_file_close (open.c:247) > ==13523== by 0x41117F5: svn_io_file_close (io.c:3471) > ==13523== by 0x50C885F: close_pack_context (pack.c:357) > ==13523== by 0x50CD2BF: pack_log_addressed (pack.c:2044) > ==13523== by 0x50CD6A3: pack_rev_shard (pack.c:2149) > ==13523== by 0x50CD892: pack_shard (pack.c:2204) > ==13523== by 0x50CDE78: pack_body (pack.c:2333) > ==13523== Address 0x69a6d04 is 68 bytes inside a block of size 4,096 > alloc'd > ==13523== at 0x402C1F0: malloc (vg_replace_malloc.c:291) > ==13523== by 0x4A6DEEB: pool_alloc (apr_pools.c:1463) > ==13523== by 0x4A6E067: apr_palloc_debug (apr_pools.c:1504) > ==13523== by 0x4A69573: apr_file_open (open.c:211) > ==13523== by 0x410BC09: file_open (io.c:351) > ==13523== by 0x411163D: svn_io_file_open (io.c:3428) > ==13523== by 0x50BA367: svn_fs_x__l2p_proto_index_open (index.c:415) > ==13523== by 0x50C827C: initialize_pack_context (pack.c:266) > ==13523== by 0x50CCFDC: pack_log_addressed (pack.c:1996) > ==13523== by 0x50CD6A3: pack_rev_shard (pack.c:2149) > ==13523== by 0x50CD892: pack_shard (pack.c:2204) > ==13523== by 0x50CDE78: pack_body (pack.c:2333) > I've got a slightly different stack but it seems to be the same problem. Took me a while to trace down the root cause. r1573237 should fix it. -- Stefan^2.
