<div class="gmail_quote">2010/3/28 Hendrik Sattler <span dir="ltr"><<a href="mailto:post@hendrik-sattler.de">post@hendrik-sattler.de</a>></span><br><blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">
But if you ever want to leave the very narrow path of what qmake is supposed<br>
to do, you are virtually doomed :-(<br></blockquote><div> </div><div>Correct, hence my switcheroo to CMake (and of course the colourful and clean makefile build has something to do with it :) <br><br></div><blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">
"gcc -s" is only convenient for e.g. qmake because it only has one install<br>
target. In cmake, you have "make install" and "make install/strip", so you can<br>
decide at a later point. And even later, you can always use strip directly to<br>
achieve exactly the same. Choosing -O2 would really be preferred, but you may<br>
as well argue for -Os or anything else. You should always set the preferred<br>
thing yourself, see (C|CPP|CXX|LD)FLAGS environment variables.<br><div class="im"></div></blockquote><div><br>Agreed on the -s part, but I still don't like a standard -O3 that, according to documentation, make your built binary perform worse. <br>
<br></div><blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">No, because when compiling differently for Debug and Release, you have all<br>
object files twice. Thus you would enforce one more level of subfolders even<br>
for those that only want one setup.<br>
There may be a new Generator that does that but it still has to be written ;)<br>
</blockquote></div><br>I agree the additional subfolder level may be unwanted, that's why I suggested the debug object file suffix to work around / solve the problem. IMHO that's a clean way of seperating the debug vs release object files.<br>
<br>And to Eric: I consider an extra script along with CMake's part in this superfluous. Why can't CMake mimic that behavior and let a user/dev use make (dist)clean? And about the multiconfig single build: yes indeed, for larger projects like WxWidgets or Qt this would be a requirement, as often devs will develop with a debug build (for obvious reasons) and then use the release build for their... release.<br>
And yes, the CMAKE_DEBUG_POSTFIX is a good thing and I use it.<br>