serf 1.3.8 test failure on VS 2015 (64-bit)

[luke...@gmx.de]

Hi,

trying to build Serf 1.3.8 on VS 2015 (64-bit) I'm running into a test failure:

1) test_deflate_buckets: test\test_buckets.c:1363: expected <0> but was <20014>

I couldn't find any references on the web so far. Anybody has an idea what might be wrong here?

Regards,
Stefan

[Lieven Govaerts]

Hi Stefan,

The test runs fine on my 64-bit Mac.

The error happens while setting up the test, it comes from compressing
1000x the same 50 character string.
It's kinda an artificial scenario, but it should work.

Maybe you can try printing out zerr where deflate_compress() return
APR_GENERAL (code 20014) ?

regards,

Lieven


> Regards,
> Stefan

[Stefan]

Hi Lieven

Trying to debug the issue turns out to be quite a challenge. When I try
to attach the debugger at the point of the error the whole system
becomes unstable (monitors turn black and graphics card is reset -
afterwards other applications start crashing (like thunderbird of firefox).

I also see that at this state test_all.exe is using > 2GB of memory.

Adding some output to the tests succeeded however. zerr is
Z_STREAM_ERROR in this case.

If you have any further idea how to trace down the root-cause it would
be welcome.

Regards,
Stefan

[Lieven Govaerts]

On Tue, Aug 11, 2015 at 1:34 PM, Stefan <luke...@gmx.de> wrote:
> Hi Lieven
>>
>> On Fri, Aug 7, 2015 at 12:23 PM, <luke...@gmx.de> wrote:
>>>
>>> Hi,
>>>
>>> trying to build Serf 1.3.8 on VS 2015 (64-bit) I'm running into a test
>>> failure:
>>>
>>> 1) test_deflate_buckets: test\test_buckets.c:1363: expected <0> but was
>>> <20014>
>>>
>>> I couldn't find any references on the web so far. Anybody has an idea
>>> what
>>> might be wrong here?
>>
>> The test runs fine on my 64-bit Mac.
>>
>> The error happens while setting up the test, it comes from compressing
>> 1000x the same 50 character string.
>> It's kinda an artificial scenario, but it should work.
>>
>> Maybe you can try printing out zerr where deflate_compress() return
>> APR_GENERAL (code 20014) ?
>
> Trying to debug the issue turns out to be quite a challenge. When I try to
> attach the debugger at the point of the error the whole system becomes
> unstable (monitors turn black and graphics card is reset - afterwards other
> applications start crashing (like thunderbird of firefox).
>
> I also see that at this state test_all.exe is using > 2GB of memory.

That unbounded memory usage has been solved by Ivan on serf trunk in
r2445. With this patch applied to 1.3.x memory usage drops from 2GB to
200MB after running the test_deflate_buckets test.

> Adding some output to the tests succeeded however. zerr is Z_STREAM_ERROR in
> this case.
>
> If you have any further idea how to trace down the root-cause it would be
> welcome.

Can you check if r2445 solves this problem for you?
thanks,
Lieven

>
> Regards,
> Stefan
>
> --
> You received this message because you are subscribed to the Google Groups
> "Serf Development List" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to serf-dev+u...@googlegroups.com.
> To post to this group, send email to serf...@googlegroups.com.
> Visit this group at http://groups.google.com/group/serf-dev.
> For more options, visit https://groups.google.com/d/optout.

[Stefan]

Hi Lieven,

>> Adding some output to the tests succeeded however. zerr is Z_STREAM_ERROR in
>>> this case.
>>>
>>> If you have any further idea how to trace down the root-cause it would be
>>> welcome.
> Can you check if r2445 solves this problem for you?

Yes, this really solves the problem here. Thanks for the hint (maybe
worth pushing a new 1.3-version with that fix included at some point?).

FYI I was able to trace down the issue to the apr_palloc-call which
returned a null-ptr in my case in deflate_compress(). Didn't dig down
deeper why it did not return any more memory here, but because write_buf
was null, the following deflate-call returned Z_STREAM_ERROR.

Regards,
Stefan