Clearing the last few regressions ????
Looks to me like we're almost there, even if it doesn't look like it yet :-) Date-Time: ~~~~~~~~~~ Serialisation lib fails to build with Borland-5.8.2: this works or me when building against spirit-1.6.3. The error messages in the table correspond to what I see when building against CVS spirit (which is known not to work). Since Borland can't possibly work with mainline Spirit, I have: using borland : 5.8.2 : "C:/Program Files/Borland/BDS/4.0/Bin/bcc32.exe" : <cxxflags>-IC:/data/boost/spirit-1.6.3/boost ; using borland : 5.6.4 : "C:/cpp/Borland/CBuilder6/Bin/bcc32.exe" : <cxxflags>-IC:/data/boost/spirit-1.6.3/boost ; In my user-config.jam which takes care of serialisation at least. Dynamic Bitset: ~~~~~~~~~~~~~~~ Strange looking failure from one regression runner, but not the other. Spurious? MultiIndex: ~~~~~~~~~~~ Spirit setup issue again. Regex: ~~~~~~ These all pass for me here. The msvc-8 failure is only showing up on one machine and the link leads to no error messages. The Borland failures all suggest that the library failed to build, but there are no error messages and since all the other tests pass there is something strange going on here (as the library must have built!). Spurious? Serialisation: ~~~~~~~~~~~~~~ Spirit setup again. Test: ~~~~~ These all pass for me, but I see an "asynch-exceptions-on" in the test name, does this mean this is a non-default build? Wave: ~~~~~ We need a Tru64 expert to look at this: linker bug maybe? John.
John Maddock <john <at> johnmaddock.co.uk> writes: [...]
Regex: ~~~~~~
These all pass for me here. The msvc-8 failure is only showing up on one machine and the link leads to no error messages. The Borland failures all suggest that the library failed to build, but there are no error messages and since all the other tests pass there is something strange going on here (as the library must have built!). Spurious?
Serialisation: ~~~~~~~~~~~~~~
Spirit setup again.
Leaving the Spirit issue aside, there is a problem I described elsewhere whereby several static libraries fail to build with the Borland 5.8.2 compiler since the switch to Boost.Build v2 . I only became fully aware of this very recently: as it didn't show up on the regression tests I had attributed it to quirks of my own setup. Moreover as I used to look for build errors by grep'ing for '[Ee]rror' in the bjam output I missed the fact that this actually doesn't involve just testing, but also building the libraries. I hope to be able to look into this as soon as the Spirit issues are solved. Cheers, Nicola Musatti
Nicola Musatti wrote:
John Maddock <john <at> johnmaddock.co.uk> writes: [...]
Regex: ~~~~~~
These all pass for me here. The msvc-8 failure is only showing up on one machine and the link leads to no error messages. The Borland failures all suggest that the library failed to build, but there are no error messages and since all the other tests pass there is something strange going on here (as the library must have built!). Spurious?
Serialisation: ~~~~~~~~~~~~~~
Spirit setup again.
Leaving the Spirit issue aside, there is a problem I described elsewhere whereby several static libraries fail to build with the Borland 5.8.2 compiler since the switch to Boost.Build v2 . I only became fully aware of this very recently: as it didn't show up on the regression tests I had attributed it to quirks of my own setup.
I know what's wrong, though I though I had fixed that very long time ago. It's on my todo for today, but later... - Volodya
Nicola Musatti wrote:
Leaving the Spirit issue aside, there is a problem I described elsewhere whereby several static libraries fail to build with the Borland 5.8.2 compiler since the switch to Boost.Build v2 . I only became fully aware of this very recently: as it didn't show up on the regression tests I had attributed it to quirks of my own setup. Moreover as I used to look for build errors by grep'ing for '[Ee]rror' in the bjam output I missed the fact that this actually doesn't involve just testing, but also building the libraries.
:-( I'd offer to help, but I can't reproduce the issue. Only thing I'm aware of that's changed, is that in bbv1 we often had to increase the library page size with Borland in order to get the static lib built. I notice a warning in the output: Warning: 'g:\boost\rc-1-34\results\boost\bin.v2\libs\regex\build\borland-5.8.2\debug\link-static\threading-multi\usinstances.obj' file not found Warning: library was too large for page size, rebuilt with page size 64 Warning: DOS-reported error: Permission denied results are safe in file 'g:\boost\rc-1-34\results\boost\bin.v2\libs\regex\build\borland-5.8.2\debug\link-static\threading-multi\libtmp00.i10' You could try adding a /P64 switch to the tlib command and see if that fixes things (but I'm not sure how to inject that with bbv2). The "DOS-reporting error" is strange to say the least as well! Not sure if this helps, John.
John Maddock wrote:
Looks to me like we're almost there, even if it doesn't look like it yet :-)
Yes. As you might have noted Nicola, Volodya and me are working on this.
Since Borland can't possibly work with mainline Spirit, I have:
using borland : 5.8.2 : "C:/Program Files/Borland/BDS/4.0/Bin/bcc32.exe" : <cxxflags>-IC:/data/boost/spirit-1.6.3/boost ; using borland : 5.6.4 : "C:/cpp/Borland/CBuilder6/Bin/bcc32.exe" : <cxxflags>-IC:/data/boost/spirit-1.6.3/boost ;
Thank you! obviously I was missing the trailing /boost on the include path! Compile looks much better this way. However we are currently trying to get the include flags into the Jamfile of serialization. I think the same should be done with DateTime. But given that spirit isn't supported at all in later versions of the compiler, your suggestion to put this in user-config.jam might be more sensible. @volodya: Are you listening? Roland
Roland Schwarz wrote:
John Maddock wrote:
Looks to me like we're almost there, even if it doesn't look like it yet :-)
Yes. As you might have noted Nicola, Volodya and me are working on this.
Since Borland can't possibly work with mainline Spirit, I have:
using borland : 5.8.2 : "C:/Program Files/Borland/BDS/4.0/Bin/bcc32.exe" : <cxxflags>-IC:/data/boost/spirit-1.6.3/boost ; using borland : 5.6.4 : "C:/cpp/Borland/CBuilder6/Bin/bcc32.exe" : <cxxflags>-IC:/data/boost/spirit-1.6.3/boost ;
Thank you! obviously I was missing the trailing /boost on the include path! Compile looks much better this way.
However we are currently trying to get the include flags into the Jamfile of serialization. I think the same should be done with DateTime.
But given that spirit isn't supported at all in later versions of the compiler, your suggestion to put this in user-config.jam might be more sensible.
@volodya: Are you listening?
Yes, I think it's more important to get regression tests running than doing it the right way. So, if changing user-config.jam as above works for you, please run tests with such change. - Volodya
Roland Schwarz wrote:
John Maddock wrote:
Looks to me like we're almost there, even if it doesn't look like it yet :-)
Yes. As you might have noted Nicola, Volodya and me are working on this.
Yep, just wasn't sure how far you had got :-)
Thank you! obviously I was missing the trailing /boost on the include path! Compile looks much better this way.
Took me about three attempts to get it right as well!
However we are currently trying to get the include flags into the Jamfile of serialization. I think the same should be done with DateTime.
But given that spirit isn't supported at all in later versions of the compiler, your suggestion to put this in user-config.jam might be more sensible.
I think the priority is to get an accurate view of what our regressions actually are, fixing the Jamfile can come later. John.
participants (4)
-
John Maddock -
Nicola Musatti -
Roland Schwarz -
Vladimir Prus