[tptacek]

Holmes ran that business in high school. She was born in 1984, so that would put the date around 2000. There were good reasons not to use G++ in 2000. When Alexandrescu's book was published, in 2001, it was somewhat notorious on my team (shipping code on G++) for how much of it didn't work well, or at all, in G++.

[saidajigumi]

Heck, virtually any C++ compiler was a big pile of sharp edges back then, by comparison to just a few years later. Especially if you dared to use it for embedded work, not so much due to the "embedded" part but just because the embedded suite compilers were just so much further behind the C++ curve.

[tptacek]

Yep, but the commercial compilers generally had better STLs (particularly for debugging), debuggers that could properly mangle/demangle symbols, and precompiled headers --- in addition to different sets of C++ features that did or didn't work properly.

[saidajigumi]

Largely true, but I recall actually isolating one chunk of code and compiling with g++ just to get halfway sane error messaging around a (IIRC) template-related error. I.e. you could do worse. The primary compiler for that code was the circa-2001 ARM tools suite. I wrote a few novels worth of feedback and bug reporting to them...

[dalke]

And to think I was shipping C++ software compiled during the egcs days. Of course, it compiled on IRIX, AIX, HP-UX and a few other compilers - I think some were still based on cfront - so we were using the boring, simple parts of C++ that everyone supported.