eBay architecture slides from 2006: https://www.cs.cornell.edu/courses/cs330/2007fa/slides/eBayS...
3.3M LoC C++, that must have been quite painful.
eBay architecture slides from 2006: https://www.cs.cornell.edu/courses/cs330/2007fa/slides/eBayS...
3.3M LoC C++, that must have been quite painful.
They replaced it with Java, which was probably worse. :p
No. At least, not nearly as many footguns. Also, compiler error messages that actually can be deciphered.
C++ compiler errors back in C++99/01 days weren't usually so bad unless you were doing basic code and not some SFINAE shit. Increased boost usage and C++11 introduced var-arg templates and a lot of other things that made things explode.
Today we've removed a lot of SFINAE magic since if-constexpr that is easier to read/trace and also the existence of CONCEPTS to pre-constrain to compatible types and thus avoiding implementation details from becoming the error.
True, on the other hand writing portable code was a mess, even if we constrained ourselves to UNIX world, as each vendor compiler supported a different flavour of the standard.
I had some fun between HP-UX aC, AIX xlC, Solaris cc, and Red-Hat gcc, and naturally there was MSVC as well.
We had #ifdefs in place for prototypes, because the version of the aC compiler we had on our development server still only supported K&R C function definitions.
Not really, there is a reason why the Java pivot into the server during the early's 2000's, it wasn't only Sun, IBM, Oracle, Bea marketing, it was definitly much better experience than mod_tcl or mod_perl without JTI[0], CGI/ISAPI in C or C++.
Likewise with ASP.NET on Windows land, as ASP with VB, and C++ alongside COM wasn't that great either.
[0] - By the time this started to matter Java 1.3 was already the common version.