Re:[boost] question/problem with process_jam_log

From: Robert Ramey (ramey_at_[hidden])
Date: 2003-12-22 15:30:57

while investigating this problem, I came upon the regress/test which tests
the regression testing system.

Great

I ran the test of the test and got the results below

$ ./test -sTOOLS=vc7
Begin test processing...
Begin log processing...
Usage: bjam [bjam-args] | process_jam_log [--echo] [locate-root]
  locate-root is the same as the bjam ALL_LOCATE_TARGET
  parameter, if any. Default is boost-root.
boost_root: C:/boost-dev/tools/regression/test/../../..
locate_root: C:/boost-dev/tools/regression/test/../../..
no errors detected
Begin compiler status processing...

**** exception(205): std::exception: boost::filesystem::directory_iterator const
ructor: "test_status.html\s\regression\test\bin": The system cannot find the pat
h specified.
******** errors detected; see standard output for details ********
The system cannot find the file test_status.html.
$ pwd
C:/boost-dev/tools/regression/test
$ ls
CVS                compile-fail.cpp   run-note-pass.cpp  run-pass.cpp
Jamfile            run-fail.cpp       run-note-warn.cpp  test.bat
bjam.log           run-note-fail.cpp  run-pass-warn.cpp  test_links.html
$ cat test.bat
@echo off
echo Begin test processing...
bjam --dump-tests %* >bjam.log 2>&1
echo Begin log processing...
process_jam_log <bjam.log
start bjam.log
echo Begin compiler status processing...
compiler_status %BOOST_ROOT% test_status.html test_links.html
start test_status.html
$

the xml file looks like:

<test-log library="" test-name="run-fail" test-type="run" test-program="tools/regression/test/run-fail.cpp" target-directory="bin/boost/tools/regression/test/run-fail.test/vc7/debug" toolset="vc7" show-run-output="false">
<compile result="succeed" timestamp="2003-12-22 20:13:02 UTC">
run-fail.cpp
</compile>
<link result="succeed" timestamp="2003-12-22 20:13:02 UTC"></link>
<run result="succeed" timestamp="2003-12-22 20:13:02 UTC">
example of output from a run-time failure
</run>
</test-log>

which is simlar to the issue I raised - result=succeed even though the test fails

Here is the relvent? part of the bjam.log which clearly shows run-fail test showing **passed.

execute-test ..\..\..\bin\boost\tools\regression\test\run-fail.test\vc7\debug\run-fail.run
====== BEGIN OUTPUT ====== 
example of output from a run-time failure
====== END OUTPUT ======  
**passed** ..\..\..\bin\boost\tools\regression\test\run-fail.test\vc7\debug\run-fail.test

Aside from my concerns regarding my particular library? is it conceivable that the status tables
are showing "passed" for tests that are failing ?  This would be a big problem.

Robert Ramey

at runtime?