So I managed to figure out the problem. It was a socket leak, caused by me
failing to destroy a pool inside a function called in a loop.
My apologies for disturbing people.
On 2012-11-27 9:33 AM, "Robert Turner" <rturner_at_qnx.com> wrote:
>Thanks for the reply.
>So I spent ages digging into this (and related issues) last night, and
>that is correct. gethostbyname() is indeed failing, spontaneously it
>seems. However, as to why, I don't know yet. I'm suspecting something is
>getting messed up in the local memory space that is affecting it. Other
>processes do not seem to be affected.
>I also tried using 1.7.7 client libraries (which I eventually got to
>compile and more or less work from source once I stopped using a custom
>built APR and used the one Apple provides on the platform). However, I was
>running into stack corruption issues during repeated svn_client_cat2()
>calls (the revision was getting overwritten, or the pointer was getting
>changed and ended up being the URL string). I haven't yet figured these
>issues out, or tried more current code than 1.7.7.
>On 2012-11-27 9:14 AM, "Daniel Shahaf" <d.s_at_daniel.shahaf.name> wrote:
>>Robert Turner wrote on Mon, Nov 26, 2012 at 19:49:35 +0000:
>>> I'm in the process of making some updates to a FUSE svnfs, and while
>>>the process is merrily chugging away pulling file information, it starts
>>>getting lots of errors from the svn_client commands (not just the one
>>>below, but that's an example):
>>> DEBUG :  svnclient_cache(): svn_client_cat(), err->message='OPTIONS
>>>of '<<<pathname removed>>>': Could not resolve hostname `<<<host name
>>>removed>>>': Host not found (<<<host name removed>>>)'
>>>err->apr_err=175002,0x2ab9a 'RA layer request failed'
>>> There is nothing of note configured related to proxies, or anything. As
>>>mentioned above, the operations are performing correctly for a while,
>>>then stop, and no further operations succeed until the process exits and
>>>is restarted. My process seems to behaving well from a memory
>>>perspective, and static analysis tools are coming up clean.
>>> I'm suspecting that something is getting confused in the internal state
>>>of the SVN client libraries.
>>And I'm suspecting your C library DNS resolver function is returning an
>>error. Have you ruled that out yet?
>>> Any suggestions on where to look further, or what might be the problem?
>>> (I'm using the 1.6.18 client libraries from wanDisco, compiled for OSX
>>>-- I have been unable to get locally compiled and working copies of the
>>>SVN libraries myself when trying to compile from source, so I have no
>>>easy way to debug through the SVN libraries)
>>> Thanks for any ideas anyone might have,
Received on 2012-12-11 02:43:29 CET