apache: strip extra dirs. #14823
apache: strip extra dirs. #14823
Conversation
|
By analyzing the blame information on this pull request, we identified @edolstra, @lethalman and @peti to be potential reviewers |
|
There are a lot of seemingly unrelated changes in this PR? |
|
Looks to me like what this PR actually does is change |
layus
force-pushed
the
closure-size-apache
branch
from
83b8783 to
7b56559
Compare
|
Yes, indeed. I got my commits messed up. |
layus
force-pushed
the
closure-size-apache
branch
from
7b56559 to
ca03137
Compare
layus
changed the title
| @@ -23,6 +23,7 @@ stdenv.mkDerivation rec { | |||
| # FIXME: -dev depends on -doc | |||
| outputs = [ "dev" "out" "doc" ]; | |||
| setOutputFlags = false; # it would move $out/modules, etc. | |||
| stripDebugList= [ "bin" "lib" "modules" ]; # also strip modules. | |||
There was a problem hiding this comment.
What is the purpose of this modification? This doesn't seem to change anything, really. Am I missing something?
There was a problem hiding this comment.
No, it does not change anything to apache 2.2, it simply unifies the syntax with 2.4.
There was a problem hiding this comment.
I could of course unify the other way around, but I feel this makes more sense with the name of the variable, and with the other pull requests I just made.
|
In fact, I am not sure this should be merged. Our stripping strategy should be smarter that to try only bin/ and lib/ and just strip any ELF in the derivation. |
|
The only problem with stripping everything is that it will give a lot of pointless warnings about non-ELF files. That could be solved by using the |
|
@edolstra This seems something we should really do. Wild references are a pain to debug. |
|
Do note though that there are other things than ELFs that presumably need stripping as well, such as |
|
If error message are really an issue, why not strip everything and &>/dev/null the output ? |
|
Also, #5447 is related. |
|
Piping to /dev/null is not a good idea because it will hide actual error messages. |
dezgeg
added
0.kind: enhancement
|
What is the state of this PR? Is it still relevant? |
|
It is still relevant as the problem still exists. However, the actual fix is being developed in #15339. That fix is taking a while to get merged as it changes stdenv (modifies the strip phase). |
We did this for 2.2 (cc61d31) but lost this for 2.4. This reduces the Apache closure size from 312 MiB to 102 MiB (primarily by getting rid of -dev outputs).
Apache keeps reference to build-time paths in /include files.
I bet these files could be simply removed, but creating a dev output works just as well.