forked from gcc-mirror/gcc
-
Notifications
You must be signed in to change notification settings - Fork 0
/
ISSUES
37 lines (28 loc) · 1.58 KB
/
ISSUES
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
=========================================
Major repo-wide problems are listed here.
=========================================
Using the configure options in configure-opts, building and installing the
kludged gcc 3.2.3 works in a single make invocation e.g. make bootstrap
install. make bootstrap then a separate make install tries to rebuild cc1,
and gets it wrong.
This might just be me not using gcc's configure and make right, but vanilla
(non-kludged) 6.3.0 doesn't have this problem.
-----
There are still 21 unexpected fails from make check-gcc on kludged 3.2.3
I expect to be backporting some things from trunk soon.
-----
On a related note, a patch in the 3.2 kludges which fixes header search dirs
is a poor, non-portable workaround for the lack of --enable-multiarch, which
existed from 2012. Anyone who was watching the repo on Github (nobody?) will
have seen it apparently dead.
Other things have taken more of Chuno's time than gcc-kludges lately. There
was a hare-brained scheme to backport multilib and multiarch support to 3.2,
which was actually working (when it stopped, the changes were successfully
applied to a point somewhere around June 2003) but also a really dumb thing to
do.
That does not mean however, that having gcc's search paths automagically just
be correct for modern versions of supported OSes is abandoned; there is a
better way to ensure it works across all of them.
However gcc-kludges stuff in general is still on the back-burner at the time
of writing (Sat 31 Aug 2019). make check-gcc will fully work without the user
having to worry about anything, one day.