Hello
I have tried to open
an archive in text mode with no_codecvt attribute,
like this:
std::ifstream ifs("archive.mob");
boost::archive::text_iarchive ia(ifs,
boost::archive::no_codecvt);
but it still triggers
the bug.
I do not think the
bug is due to something that is not handled in the stream, since I do not have
any special characters or stuff like that (except end lines). Plus, the fact
that it crashes at the magical number 512 characters get me to think it is
more related to a buffer problem, however, you are the
specialist.
How can I investigate
the code conversion facets used by text archive? I am not familiar at
all with locales, or facets, but if you give me some hints about how they are
used in the BSL, I may get started.
Regards
Jean-Noël
Ps: everything works
fine in XML archive mode, so for my project I have switched to that, but I’d
still like to get this bug fixed for the benefits of
all…
De :
boost-users-bounces@lists.boost.org
[mailto:boost-users-bounces@lists.boost.org] De la part de Robert Ramey
Envoyé : jeudi 9 février 2006
17:11
À :
boost-users@lists.boost.org
Objet : Re: [Boost-users]
[Serialization]: bug while
serializingastd::string
Sorry I got mixed up with another
post.
The main difference between
strstream and other streams is that strstream doesn't do any code
conversion. This is also consistent with the idea that something in the
string isn't being handled in the code conversion facet of the the stream in
the expected manner. a couple of ideas would be to investicate the code
conversion facets used by the text archive. Also look at the no_codecvt
attribute on archive open.
I am assuming you
meant :
std::ofstream ifs("container.mob");
boost::archive::text_oarchive
ia(ofs);
oa
<< MyObject;
ofs <<
std::endl;
and not
std::ifstream ifs.
So if I understand well,
this “patch” only adds a newline to the archive at the saving
stage.
I tried it but it did not
help in any way; in fact I cannot see how it could help, since the problem
is at the point of the loading of the std::string, not at the end of the
load of the whole object. What I mean here is that MyObject is, in my case,
a class containing a std::vector of pointers to class containing
std::strings. One of these strings is too long, which causes the crash.
Maybe what you meant is appending the std::endl after the serialization of
the std::string, but this is not possible since at this point I have only
access to the archive, not the underlying ofstream
:
template<class
Archive>
void
serialize(Archive & ar, const MyClass& m, const unsigned int
version)
{
ar & m.a_standard_string;
}
I do not have
access there to the underlying ofstream of the implementation. Note that
here, MyObject would contain a std::vector < MyClass * >. And I am
serializing directly MyObject which causes the serialization of MyClass
later.
So, what do you
mean? Have you located the bug ?
De :
boost-users-bounces@lists.boost.org
[mailto:boost-users-bounces@lists.boost.org] De la part de Robert Ramey
Envoyé : jeudi 9 février 2006
04:46
À :
boost-users@lists.boost.org
Objet : Re: [Boost-users]
[Serialization]: bug while serializing
astd:string
Untl 1.34 comes out try the
following on the output stream
std::ifstream ofs("container.mob");
boost::archive::text_oarchive
ia(ofs);
oa
<< MyObject;
ofs <<
std::endl;
I have been able to test
with a string stream; the bug does not occur.
That is, this code
produces the bug:
std::ifstream ifs("container.mob");
boost::archive::text_iarchive
ia(ifs);
ia
& MyObject;
while the following does
not:
std::ifstream ifs("container.mob");
std::ostringstream tmp;
tmp <<
ifs.rdbuf();
std::istringstream is(tmp.str());
boost::archive::text_iarchive
ia(is);
ia
& MyObject;
a)
try using a different stream
buffer –
strstream
So do you have
any clue, Robert ?? Clearly the problem has to do with buffering
somewhere….
Jean-Noël
_______________________________________________
Boost-users
mailing
list
Boost-users@lists.boost.org
http://lists.boost.org/mailman/listinfo.cgi/boost-users
_______________________________________________
Boost-users
mailing
list
Boost-users@lists.boost.org
http://lists.boost.org/mailman/listinfo.cgi/boost-users