Test case failures on OSX

[Mark Treiber-2]

I've finally been able to get around to testing the one of the 0.3.8
gm releases and I had two test case failures on OSX.  I wanted to
check to see if others are getting the same failures on other
platforms or not before investigating further.  One of the failures
seems to be a precision based failure (this was the case for a test
case failure in 0.3.6 that was fixed) but the other one seems
universal.  The test case output is below.

Running 172 test cases...
bonds.cpp(182): fatal error in "BondTest::testCachedYield": failed to
reproduce cached price:
    calculated: 99.223769
    expected:   99.204505
    error:      0.019264
Exception in "CapFloorTest::testCachedValue": std::exception:
indexmanager.cpp:30: In function `const QuantLib::History&
QuantLib::IndexManager::getHistory(const std::string&) const':
Euribor6m act/360 history not loaded
Exception in "MoneyTest::testBaseCurrency": std::exception:
exchangeratemanager.cpp:170: In function `QuantLib::ExchangeRate
QuantLib::ExchangeRateManager::smartLookup(const QuantLib::Currency&,
const QuantLib::Currency&, const QuantLib::Date&, std::list<Integer,
std::allocator<Integer> >) const':
no conversion available from GBP to EUR for December 10th, 2004
Exception in "MoneyTest::testAutomated": std::exception:
exchangeratemanager.cpp:170: In function `QuantLib::ExchangeRate
QuantLib::ExchangeRateManager::smartLookup(const QuantLib::Currency&,
const QuantLib::Currency&, const QuantLib::Date&, std::list<Integer,
std::allocator<Integer> >) const':
no conversion available from EUR to GBP for December 10th, 2004
piecewiseflatforward.cpp(228): fatal error in
"PiecewiseFlatForwardTest::testObservability": Observer was not
notified of date change
Exception in "SwapTest::testCachedValue": std::exception:
indexmanager.cpp:30: In function `const QuantLib::History&
QuantLib::IndexManager::getHistory(const std::string&) const':
Euribor6m act/360 history not loaded

These were all run with the Dec 9 gm source dist.

Mark.

[Luigi Ballabio-2]

Hi Mark,
        thanks for checking this.

On 2004.12.11 00:34, Mark Treiber wrote:
> I've finally been able to get around to testing the one of the 0.3.8
> gm releases and I had two test case failures on OSX.  I wanted to
> check to see if others are getting the same failures on other
> platforms or not before investigating further.

No, that's the first I heard of such failures. As far as I know, a number  
of compilers on Windows and gcc on Debian x86 all pass the check.

> One of the failures seems to be a precision based failure (this was the  
> case for a test case failure in 0.3.6 that was fixed)

What was the fix?

> but the other one seems universal.  The test case output is below.

[snip]

Hmm, I might have a suspect. Can you try the attached patch on ql/Patterns/
singleton.hpp and see if this fixes the problem?

Later,
        Luigi