
Hi Robert, I have seen that you have changed the libs/serialization/build/Jamfile.v2 Some comments on this: 1) The issue not only applies to borland, and not only to serialization. 2) I am running the regressions with a modification to user-config.jam. I.e.: using borland : : : <cxxflags>-Ipath_to_spirit ; This works well, and if the path is not set the compiler will spill out a message, that will inform the user, that another version of spirit is needed. 3) Your change will insist on a SPIRIT_ROOT beeing available, despite the fact that the include already might have been specified correctly for the toolset. 4) Will your change also work with path-constant, set in site-config.jam ? Policy is to get away with environment variables. Thank you, Roland

Roland Schwarz wrote:
Hi Robert,
I have seen that you have changed the libs/serialization/build/Jamfile.v2
Some comments on this:
1) The issue not only applies to borland, and not only to serialization.
2) I am running the regressions with a modification to user-config.jam. I.e.:
using borland : : : <cxxflags>-Ipath_to_spirit ;
This works well, and if the path is not set the compiler will spill out a message, that will inform the user, that another version of spirit is needed.
I personally think that adding extra include in toolset configuration is the most clean approach. We need not be extra smart for support of non-mainstream configurations.
3) Your change will insist on a SPIRIT_ROOT beeing available, despite the fact that the include already might have been specified correctly for the toolset.
4) Will your change also work with path-constant, set in site-config.jam ? Policy is to get away with environment variables.
Robert, for avoidance of doubt, whatever the outcome of this thread might be, may I ask you not to merge this change to RC branch? We have regression tests running OK as it is now, so let's not touch that. - Volodya

I have not merged in any changes into the RC branch for many months. (err - except for one accident). All my changes are in the HEAD. Robert Ramey
Robert, for avoidance of doubt, whatever the outcome of this thread might be, may I ask you not to merge this change to RC branch? We have regression tests running OK as it is now, so let's not touch that.
Amen

I've been happy to let you guys address the borland/spirit issue as you're clearly in a better position to address it than I. So feel free to fiddle with this stuff. If you don't have cvs access, just email me your Jamfile.v2 and I'll check it in. Ok, we'll try. Yes I do have CVS write access.
Ultimately, I would like to see the Jamfile.v2 detect that spirit 1.6x is not available and then skip the build of the xml_?archives (after emitting a warning). The same would be applied to the tests. In this way, the issue wouldn't have be dealt with for those that people that don't care about xml archives. Understood. I have seens this from the former v1 file that was included into
Robert Ramey wrote: the Jamfile.
The same goes for systems that don't support wchar_t. I would like to see wserialization and the associated tests skipped for systems which don't support wide characters.
Ok. I see. Could this be the case for the stlport? There are a lot of errors, possibly related to this issue.
Roland Schwarz wrote:
Hi Robert,
I have seen that you have changed the libs/serialization/build/Jamfile.v2
Some comments on this:
1) The issue not only applies to borland, and not only to serialization.
2) I am running the regressions with a modification to user-config.jam. I.e.:
using borland : : : <cxxflags>-Ipath_to_spirit ;
This works well, and if the path is not set the compiler will spill out a message, that will inform the user, that another version of spirit is needed.
3) Your change will insist on a SPIRIT_ROOT beeing available, despite the fact that the include already might have been specified correctly for the toolset.
4) Will your change also work with path-constant, set in site-config.jam ? Policy is to get away with environment variables.
Thank you, Roland _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
participants (3)
-
Robert Ramey
-
Roland Schwarz
-
Vladimir Prus