hansen@pegasus.ATT.COM (Tony L. Hansen) (08/17/90)
One of the "library" issues that Bjarne has shied away from specifying within the reference part of E&S is that of the contents of <new.h> and the guaranteed availability of set_new_handler(). Since E&S does NOT deal with library issues at all, they (perhaps rightly) did not include anything within it regarding <new.h> or set_new_handler(). Now if you look at page 280 on free store, you WILL see a description of set_new_handler() WITHIN the commentary. However, this is NOT part of the reference manual and has no binding on anything or anyone. I found little else within E&S regarding <new.h> other than a statement on page 7 requiring its existence. The exact contents of <new.h> need to be specified within the final standard, along with the expected interaction of operator new() in the presence of a new handler. (Does new() loop? It's not required anywhere!) Tony Hansen att!pegasus!hansen, attmail!tony hansen@pegasus.att.com
mjv@objects.mv.com (Michael J. Vilot) (08/21/90)
Tony Hansen writes: > ... E&S does NOT deal with library issues at all ... > The exact contents of <new.h> need to be specified within the final standard In some ways, the standardization of the ANSI C++ library is more of a challenge than the language itself. The hardest part is that the availability of templates and exceptions will eventually change what we think is reasonable in a C++ library's design. If we assume exceptions are part of the standard, we may find the _new_handler approach is inappropriate. On the other hand, if we end up with an exception specification which has undesirable characteristics when running out of free heap and/or stack space, we'll need to retain this alternate approach. Of course, we can't decide yet -- exceptions are not part of the standard, and it will be some time before we have enough feedback from whatever scheme we adopt. It's certainly appropriate that E&S does not deal with library issues -- we clearly need a lot more experience and discussion before we're in a position to ``standardize existing practice.'' -- Mike Vilot, ObjectWare Inc, Nashua NH mjv@objects.mv.com (UUCP: ...!decvax!zinn!objects!mjv)