Created by Amos Jeffries and last modified
Get this branch:
bzr branch lp:~squid/squid/qa-tests
Members of squid can upload to this branch. Log in for directions.

Branch merges

Related bugs

Related blueprints

Branch information


Recent revisions

14883. By Amos Jeffries on 2016-10-11

QA: Regenerate layer-02-maximus

14882. By Amos Jeffries on 2016-10-10

QA: Regenerate layer-01-minimal

14881. By Amos Jeffries on 2016-10-09


14880. By Amos Jeffries on 2016-10-09

Prep for 4.0.15 and 3.5.22

14879. By eduard-bagdasaryan on 2016-10-08

HTTP: MUST ignore a [revalidation] response with an older Date header.

Before this patch, Squid violated the RFC 7234 section 4 MUST
requirement: "When more than one suitable response is stored, a cache
MUST use the most recent response (as determined by the Date header
field)." This problem may be damaging in cache hierarchies where
parent caches may have different responses. Trusting the older response
may lead to excessive IMS requests, cache thrashing and other problems.

14878. By Alex Rousskov on 2016-10-07

Optimized/simplified buffering: Appending nothing is always possible.

This change avoids CoW when appending an empty buffer to MemBlob or
SBuf, making "if !empty then append()" caller complications unnecessary.

14877. By Christos Tsantilas on 2016-10-07

Do not leak Downloader-related objects.

The downloadFinished() method was responsible for the job clean up, but
that asynchronous method is not called when the Downloader job quits
before the call can be fired. This early termination happens when, for
example, the job finishes while still inside the start() method (e.g., a
memory hit with no async ACLs to check). It also happens if an exception
is thrown while the job is running under async call protections.

Ensure the cleanup happens regardless of the job path to finish.

This is a Measurement Factory project.

14876. By Alex Rousskov on 2016-10-06

Fix known "concurrent c_str()s" violations of SBuf API.

The second c_str() call destroys the buffer still being used by the
first c_str() result, leading to many "Invalid read of size N" errors.

IMO, we must instead fix SBuf to make similar violations unlikely, but
there is currently no squid-dev consensus on whether and how to do that.
See "[RFC] Support concurrent SBuf::c_str() calls" thread on squid-dev.

14875. By Amos Jeffries on 2016-10-06

Fix header name mismatching after rev.14870

When a mime header set contains two custom headers and one
name is the prefix for the other the name lookup using a
fixed length for String comparison can wrongly match the
longer header as being equal to the shorter one, since only
the identical prefix portion is compared.

To avoid this we must check that the lengths are also matching.

This also improves performance very slightly as the common
case for custom headers is to have an "X-" prefix which is
slower to compare than total length. Headers having same
length and same prefix is quite rare.

14874. By Alex Rousskov on 2016-10-06

Hide OpenSSL tricks from Valgrind far-reaching initialization errors.

This change has no effect unless ./configured --with-valgrind-debug.

OpenSSL, including its Assembly code, contains many optimizations and
timing defenses that Valgrind misinterprets as uninitialized value
usage. Most of those tricks can be disabled by #defining PURIFY when
building OpenSSL, but some are not protected with PURIFY and most
OpenSSL libraries are (and should be) built without that #define.

To make matters worse, once Valgrind misdetects uninitialized memory, it
will complain about every usage of that memory. Those complaints create
a lot of noise, complicate triage, and effectively mask true bugs.
AFAICT, they cannot be suppressed by listing the source of that memory.

For example, this OpenSSL Assembly trick:
    Uninitialised value was created by a stack allocation
       at 0x556C2F7: aesni_cbc_encrypt (aesni-x86_64.s:2081)

Triggers many false errors like this one:
    Conditional jump or move depends on uninitialised value(s)
       by 0x750838: Debug::Finish()
       by 0x942E68: Http::One::ResponseParser::parse(SBuf const&)

This change marks OpenSSL-returned decrypted bytes as initialized. This
might miss some true OpenSSL bugs, but we should focus on Squid bugs.

Branch metadata

Branch format:
Branch format 7
Repository format:
Bazaar repository format 2a (needs bzr 1.16 or later)
Stacked on:
This branch contains Public information 
Everyone can see this information.