| 1 |
Andreas K. Huettel posted on Sun, 23 Oct 2011 15:50:04 +0200 as excerpted: |
| 2 |
|
| 3 |
> I'd like to get my standards up to speed, so may I respectfully ask- |
| 4 |
> what is, |
| 5 |
> apart from link time, the Gentoo-user-visible difference between * |
| 6 |
> removing the .a files in the ebuild * and not building them in the first |
| 7 |
> place? |
| 8 |
|
| 9 |
If it was only link-time, that might be a point for simply removing them |
| 10 |
after link. |
| 11 |
|
| 12 |
But if static libs are built at all, on amd64, it's not simply the link- |
| 13 |
time at issue, but forces an actual double-build at least for the libs, |
| 14 |
once with -fPIC for dynamic linking, once without, for the static libs, |
| 15 |
(and executables if any). |
| 16 |
|
| 17 |
Ask the mysql guys about the trouble they had with mysql-embedded on |
| 18 |
amd64, and the kde guys about the trouble with amarok for kde4 on amd64, |
| 19 |
as a result. |
| 20 |
|
| 21 |
Interestingly enough, unless I've misunderstood, this issue would be |
| 22 |
affected by the recent security-based -fPIC/-fPIE on amd64 by default |
| 23 |
discussion as well, since if everything (including static libs) were |
| 24 |
built with at least -fPIC as required for dynamic linking, then a single |
| 25 |
build, linked once each for static and dynamic, as common on x86 (32-bit) |
| 26 |
would be at least arguably acceptable. (IOW, my point here isn't to |
| 27 |
argue whether that'd be acceptable or not, but rather, that the forced |
| 28 |
double-build factor on amd64 is much more an issue than simple double- |
| 29 |
linking would be.) |
| 30 |
|
| 31 |
-- |
| 32 |
Duncan - List replies preferred. No HTML msgs. |
| 33 |
"Every nonfree program has a lord, a master -- |
| 34 |
and if you use the program, he is your master." Richard Stallman |
Archives