CUPS thinks it knows better than its users and tries to force is own clflags which can cause the compile to fail, if the compile is successful the resulting binaries will segfault. On my x86 uclibc system, CUPS failes to compile. Reproducible: Always Steps to Reproduce: 1. unpack 2008 uclibc stage3 2. emerge cups Actual Results: compile fails because the mantohtml.c included in cups segfaults, if the makefiles are patched to use man2html the compile will succeed but then cupd will segfault Expected Results: cups should respect user cflags and it should build completly as expected
Created attachment 183922 [details, diff] This patch removes cups' horrible optimization attemps
(In reply to comment #1) > Created an attachment (id=183922) [edit] > This patch removes cups' horrible optimization attemps > With this patch cups compiles and runs as expected on my uclibc system.
Confirming that problem described in this issue exists. I'm on armv6j-gentoo-linux-gnueabi.
anything gonna happen with this... its been reassigned for well over a week now
Please open an upstream bug about this issue and link the URL here to get this further.
Well since they are the ones who are overriding/changing the flags that we pass them thinking they know better I really doubt will will do anything about this if I opened a bug. I tried talking to a dev on IRC and that went no where...
Ok I filed a bug report on CUPS's bugtracker.... Here is the link if you want to see it go nowhere -> http://www.cups.org:80/str.php?L3134+P0+S-2+C0+I0+E0+M20+Q
Created attachment 191277 [details] ebuild which allows custom cflags depending on use-flag added a new USE-flag "usercflags" which applies the patch when active
Comment on attachment 191277 [details] ebuild which allows custom cflags depending on use-flag Please attach a unified diff next time. Also, with the cups ./configure script supporting --with-optim=..., which actually works in the 1.3 branch, I don't see a reason not to use that instead.
Fixed in 1.3-r*. 1.4* does not seem to suffer this issue, although it does ignore LDFLAGS (which is a matter for another bug).