Stripping binaries: How much?

[LigH]

I'm using different build environments to compile video encoders, especially x265. They are: an older MSYS / MinGW32 package with GCC 4.8.2, released years ago by XhmikosR; a current MSYS / MinGW32 package with GCC 4.9.2 by XhmikosR as well; and the media-autobuild_suite by jb_alvarado, based on MSYS2 / MinGW64 with GCC 4.9.2.

The latter uses the call strip --strip-all and produces remarkably smaller executables than the results of my manual build environments, calling strip --strip-unneeded instead.

Is it safe to use strip --strip-all or do you know of cases where this would cause issues (especially with executables including many libraries, like ffmpeg)?

[LoRd_MuldeR]

https://www.technovelty.org/linux/stripping-shared-libraries.html

So, conclusions? --strip-all is safe on shared libraries, because global symbols remain in a separate section, but not on objects for inclusion in static libraries (relocatable objects). --strip-unneeded is safe for both, and automatically understands shared objects do not need any .symtab entries to function and removes them; effectively doing the same work as --strip-all. So, --strip-unneeded is essentially the only tool you need for standard stripping needs!

[LigH]

Thank you. So I'll stay on the safe side. Sounds like --strip-all might be a risk for a multilib build. Or does it only apply to stripping a linkable library, but not to stripping a final executable? ... I am not yet sure.

[LoRd_MuldeR]

Thank you. So I'll stay on the safe side. Sounds like --strip-all might be a risk for a multilib build. Or does it only apply to stripping a linkable library, but not to stripping a final executable? ... I am not yet sure.

For shared libraries, i.e. DLL files on Windows (.so files on Linux), using --strip-all is "safe" to use. But since --strip-unneeded essential does the same thing for this kind of files, there is probably not much reason to prefer --strip-all over --strip-unneeded. If you really want to be sure, you can use Dependency Walker (http://www.dependencywalker.com/) to check that non of the exported symbols have disappeared from the DLL file after the stripping ;)

It does makes a difference for static libraries, i.e. lib files on Windows (.a files on Linux) – or, more specifically, for the object (.o) files that are contained in the static library file. So, for those files, you probably don't want to use --strip-all.

(As far as executables are concerned, I don't think you need to worry, since EXE files generally don't even have any exported symbols. So the only thing to be stripped are the debug symbols)

[LigH]

Only final executables or shared libraries (EXE or DLL) are stripped.

So I wonder why the multilib EXE produced by media-autobuild_suite is remarkably smaller than the multilib EXE produced by a script very close to the build script published by Multicoreware (IIRC, about 7 MB to 9 MB). But well, there will certainly be different optimization options in the compiler calls. So it may not be caused by stripping.

[LoRd_MuldeR]

I suspect it's something else. But you can always apply strip --strip-all manually on the final DLL or EXE file in order to make sure that both variants are "fully" stripped.