[
https://issues.apache.org/jira/browse/AVRO-1846?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15288936#comment-15288936
]
Joanne Dodd commented on AVRO-1846:
-----------------------------------
Yes the problem is with the long data field (field 1 in record event). It
follows the int so is offset by 4 bytes into the buffer which is causing the
crash.
I am not going to be able to use the workaround of changing the schema.
Agree that the solution is to insert bytes where required so that fields are
aligned correctly in the buffer as required by their type.
> Memory alignment issue causing crash on Solaris on Sparc
> --------------------------------------------------------
>
> Key: AVRO-1846
> URL: https://issues.apache.org/jira/browse/AVRO-1846
> Project: Avro
> Issue Type: Bug
> Components: c
> Affects Versions: 1.8.0
> Environment: SunOS 5.10 Generic_141444-09 sun4u sparc
> SUNW,Sun-Fire-V240
> gcc version 4.9.0 (GCC)
> Reporter: Joanne Dodd
>
> Excerpt of output from running process through gdb (note debug offsets on):
> Record address
> Header: Offset 0, size 0
> Field 0:
> Schema string
> Offset 0, size 24
> Field 1:
> Schema string
> Offset 24, size 24
> Field 2:
> Schema int
> Offset 48, size 4
> Field 3:
> Schema string
> Offset 52, size 24
> Field 4:
> Schema int
> Offset 76, size 4
> TOTAL SIZE: 80
> Record Event
> Header: Offset 0, size 0
> Field 0:
> Schema int
> Offset 0, size 4
> Field 1:
> Schema long
> Offset 4, size 8
> Field 2:
> Schema long
> Offset 12, size 8
> Field 3:
> Schema SeAddress
> Offset 20, size 8
> Field 4:
> Schema array
> Offset 28, size 16
> TOTAL SIZE: 44
> Program received signal SIGSEGV, Segmentation fault.
> avro_generic_long_init (iface=0xff26e8fc <AVRO_GENERIC_LONG_CLASS>,
> vself=0x5297614) at generic.c:1469
> 1469 *self = 0;
> (gdb) p self
> $1 = (int64_t *) 0x5297614
> (gdb)
> long integers must be aligned on 64-bit boundaries for SPARC systems.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
