c++.stl.port - Regression test results
- "Chris Idzerda" <cidzerda sbcglobal.net.nospam> Nov 13 2003
Here are the results of the regression tests I compiled and ran (in stlport/test/regression). I am using DMC 8.37 (downloaded) and STLport 4.5.3 (with iostreams). The dm.mak makefile does not work with the make program included with DM. Using "make -f dm.mak" as the command line I get: Error: don't know how to make '.cpp' If instead I run "make -f dm.mak all" I get: Error: don't know how to make 'all' (I ended up using $oMeone el$e'$ make program.) The makefile has no targets for actually running the tests so I did so manually, redirecting the output of each program to a file. I then compared that file with stl_test.exp. Almost everything worked. Here is how they differed. tstdeq compiles but crashes, probably because the linker complains: OPTLINK : Warning 134: No Start Address That's absolutely true. It doesn't have the same use of the MAIN preprocessor symbol that the others have. After minor modification, it compiled, linked, and ran correctly (producing no output). stat.cpp contains nothing more than three newlines. Is this expected? It still compiles but, not surprisingly, the linker complains: OPTLINK : Warning 23: No Stack OPTLINK : Warning 134: No Start Address Nonetheless, it still produces an executable which emits garbage when run. The output of genern1 is different from that found in stl_test.exp. It's supposed to be: 2749 12767 9084 12060 32225 17543 25089 21183 25137 25566 I got: 16838 5758 10113 17515 31051 5627 23010 7419 16212 4086 This is probably irrelevant since this program uses random numbers without specifying a seed. This is true for nthelem1, nthelem2, parsrt1, parsrtc1, ptition1, rndshuf0, rndshuf1, rndshuf2, and stblptn1 as well. I recommend changing these tests to use a fixed pseudo-random sequence instead of calling rand to facilitate comparisons.
Nov 13 2003