On Sep 10, 2003, at 9:41 PM, Naveen Kumar Kaushik wrote:
You are correct but than next question is why did people face lot of
problems in compiling ethereal-0.9.14 with MSVC++ . If I remember
correctly It could only be compiled after few patches to Makefiles.
The only Windows build problems I see in the mail archives (based on
subject lines) after July 23rd are:
1) the issue about the new way of generating the "plugin API" files -
that didn't require patches to Makefiles, it could have been solved by
setting the mod times on the generated files so it didn't try to
regenerate those files, and hopefully we can put an end to the whole
"plugin API" stuff with the changes Gisle Vanem suggested;
2) Gisle's MinGW build problems, but MinGW builds are very different
from MSVC++ builds, which are what we do;
3) Pierre Pacchioni's problems, which appear largely to have been
build environment problems;
4) Bill Meier's problems, one of which was a config.nmake problem (I
don't know why that wasn't an issue when Gerald built the binaries) and
one of which was the "plugin API" build problem, which wouldn't have
been an issue for Gerald if he had all the relevant tools installed;
5) one MSVC++ 7 problem, but nobody who has MSVC++ 7 has, as far as I
know, contributed fixes, and I, at least, am loath to install MSVC++ 7,
as
1) I don't know if it works on NT 4.0, and I don't want to upgrade to
W2k until we either find somebody willing to do builds and tests on NT
4.0 or we get assurances from the entire user base that they won't
complain if Ethereal stops working on NT 4.0;
2) upgrading to a Shiny New Compiler means you run the risk of not
catching changes that keep it from compiling with the Boring Old
Compiler, which many people might still have.
Hopefully, we can fix the "plugin API" problem for good with Gisle's
changes; that one was a new problem, because the new scheme (which I
feared, apparently correctly, might cause problems for users who didn't
have Python and gcc installed) was introduced in 0.9.14. (The old
scheme had its own problems, as you had to update several files in
parallel to add a new routine to the "plugin API".)
Gisle contributed changes, as I remember, to fix at least some of the
MinGW build problems.
There are limits on what we can do about build environment problems.
The config.nmake problem is fixed.
The MSVC++ 7 problems await somebody with the time and energy to devote
to making Ethereal build on MSVC++ 7.
