| 1 |
Here's a draft of a paragraph discussing CFLAGS-related problems. This |
| 2 |
is the result of a discussion I started on gentoo-dev. Thanks to all |
| 3 |
devs who gave feedback this should bring some food for thought to GWN |
| 4 |
readers. |
| 5 |
|
| 6 |
--- Draft BEGIN --- |
| 7 |
<section> |
| 8 |
<title>CFLAGS</title> |
| 9 |
<body> |
| 10 |
|
| 11 |
<p> |
| 12 |
Being able to tune CFLAGS is part of the user control and extreme |
| 13 |
configurability that are hallmarks of the <uri |
| 14 |
link=http://www.gentoo.org/main/en/about.xml> Gentoo experience</uri>. |
| 15 |
Being in control brings both benefits and problems. CFLAGS tuning is not |
| 16 |
an exception. |
| 17 |
</p> |
| 18 |
|
| 19 |
<warn> |
| 20 |
We would like to remind you that using anything beyond <c>-O2 |
| 21 |
-fomit-frame-pointer -march/-mcpu/-mtune</c> in CFLAGS or CXXFLAGS (and |
| 22 |
<c>-mieee, -mabi</c> etc. on selected archs that tell you to do this), |
| 23 |
and using anything at all in LDFLAGS or ASFLAGS, is usually not worth it |
| 24 |
for most users. There's usually very little benefit, high risks and much |
| 25 |
time spent on frustating tuning that could be enjoyed doing far more |
| 26 |
interesting things. |
| 27 |
</warn> |
| 28 |
|
| 29 |
<p> |
| 30 |
The recent upgrade to gcc-4.1.1 for x86 and amd64 users changed the |
| 31 |
CFLAGS landscape. Users that spent some time tuning their CFLAGS with |
| 32 |
gcc-3.4.6 might find out that an upgrade to gcc-4.1.1 leaves them with |
| 33 |
an unstable system. |
| 34 |
</p> |
| 35 |
<p>Example of this are :</p> |
| 36 |
<ul> |
| 37 |
<li>nss_ldap stopped working with <c>-ffast-math</c> (<c>-ffast-math</c> |
| 38 |
is often misused and must be considered a dangerous flag)</li> |
| 39 |
<li><c>-fvisibility-inlines-hidden</c> still breaks some code</li> |
| 40 |
<li><c>-ftree-loop-linear</c> now breaks in gcc-4.1 (at least with |
| 41 |
mesa)</li> |
| 42 |
<li><c>-ftree-vectorize</c> is known to be broken in gcc-4.1 (at least |
| 43 |
for x86 and ppc, there are fewer problems reported by amd64 users but no |
| 44 |
guarantees)</li> |
| 45 |
<li><c>-fforce-addr</c> and <c>-fweb</c> break regularly on x86 with |
| 46 |
video libraries or graphic processing apps which use hand-optimised ASM. |
| 47 |
<c>-fweb</c> may be safe on amd64 but like above no guarantees</li> |
| 48 |
</ul> |
| 49 |
|
| 50 |
<p>There are known-to-be-broken flags for all GCC versions that you want |
| 51 |
to check for too:</p> |
| 52 |
<ul> |
| 53 |
<li>-fvisibility=hidden</li> |
| 54 |
<li>-frename-registers (may be safe on amd64, at your own risks)</li> |
| 55 |
<li>-ftracer</li> |
| 56 |
<li>-D_FILE_OFFSET_BITS=64</li> |
| 57 |
<li>-msse -mmmx -m3dnow (no need for them on amd64, there are wrapped up |
| 58 |
by -march=k8/nocona/... and safely used there)</li> |
| 59 |
<li>-W</li> |
| 60 |
<li>-mfpmath=sse,387</li> |
| 61 |
<li>-malign-double</li> |
| 62 |
</ul> |
| 63 |
|
| 64 |
<p> |
| 65 |
Users with unsupported CFLAGS might want to return to safe CFLAGS (see |
| 66 |
warning above) if recent updates caused them stability problems. On the |
| 67 |
other hand, more adventurous users might want to experiment with CFLAGS |
| 68 |
that didn't work properly with gcc-3.4.6... As always, the user is in |
| 69 |
control (and the gun pointed to their feet is in his/her hand). |
| 70 |
</p> |
| 71 |
|
| 72 |
<p>Final notes:</p> |
| 73 |
<ul> |
| 74 |
<li>The gcc man page contains warnings for some unsafe optimization |
| 75 |
options. You should read it carefully when you experiment with CFLAGS or |
| 76 |
upgrade GCC on a CFLAGS-customized Gentoo.</li> |
| 77 |
<li>Some options that are unsafe in the system-wide CFLAGS might be |
| 78 |
added automatically in some ebuilds if the developper deems them safe |
| 79 |
(by redefining CFLAGS or using append-flags of the flag-o-matic eclass). |
| 80 |
For example <c>-ffast-math</c> is added by the xmame/xmess ebuilds on |
| 81 |
most architectures even tho you SHOULD NOT put it in your CFLAGS.</li> |
| 82 |
<li>You might get an idea of the stability issues of a specific |
| 83 |
optimization option by running: <c>find /usr/portage -name '*.ebuild'| |
| 84 |
xargs grep -- '-<your-risky-optimization-option>'</c>. It takes quite |
| 85 |
some time, but might be enlightening: look for the |
| 86 |
'<c>filter-flags</c>'.</li> |
| 87 |
</ul> |
| 88 |
|
| 89 |
</body> |
| 90 |
</section> |
| 91 |
--- Draft END --- |
| 92 |
|
| 93 |
Lionel |
| 94 |
-- |
| 95 |
gentoo-dev@g.o mailing list |
Archives