Hello! > > Even though it is probably OK to assume that <inttypes.h> exists whenever > > <stdint.h> does, keeping using PCI_HAVE_STDINT_H for that is confusing :) > > Are you suggesting just removing the defined(PCI_HAVE_STDINT_H) part? I suggest either removing the test altogether (if we come to a conclusion that we can rely on C99 includes everywhere), or renaming the symbol to PCI_HAVE_INTTYPES_H. > Also, I looked at the configure script and it looks like > PCI_HAVE_STDINT_H is only defined on sunos. That doesn't make much > since to me. AFAIK old versions of Solaris had some C99 includes, but not a C99 compiler, so the STDC_VERSION test did not trigger. > For non-MS Windows systems, what systems do we support that don't have > those c99 headers? As I have already said, I am not sure about that as I have never seen some of the supported systems myself. Current versions of all systems except Windows should be safe, but older versions frequently lack them. > I found that AIX 4.x, which has been end of lifed for a while, doesn't > have the headers, but AIX 5.3 does. Do we really need to maintain > support of systems as old as AIX 4.x? Judging from the occasional bug reports I receive, people still use such systems, so as long as supporting the old systems is a matter of a couple of lines, I would really like not to break them. I would therefore keep the fallback to u_intXX_t types, but instead of u_int64_t (for which we do not know the format sequence) use either long or long long, depending on ULONG_MAX. > Also, do we have a list of supported platforms somewhere? They are listed in README, but only names of the platforms, not specific versions. > I don't understand this comment fully. Previously, these values were > u32. Didn't they get printed the same way, that is using the non-u32 > format strings? In exec_op(), we call: printf(formats[width]+1, x); (and similarly trace()). I.e., we are printing the same variable `x', but with different format strings. As long as the format strings were "%<something>x", it worked, because all those required `unsigned int' as an argument. If you change `x' to u64, you have to change _all_ format strings used on it. Have a nice fortnight -- Martin `MJ' Mares <mj@xxxxxx> http://mj.ucw.cz/ Faculty of Math and Physics, Charles University, Prague, Czech Rep., Earth Q: How many Prolog programmers does it take to change a light bulb? A: No. -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
