llvm dev - May 2018 - Removing LLVM testing tools from the install distribution

Should the binaries included in an LLVM release be generally useful to a wide
audience of toolchain users, or is it OK to ship tools that are only really
useful for LLVM development?

If the former is the case, can we consider not shipping tools which are
exclusively for testing the compiler? E.g removing these 5 binaries from
5.0.2-rc1 would have saved over a hundred megabytes in uncompressed space for
macOS users:

Binary name | Size in megabytes
--------------------------------------------
clang-check: 54.6725
llvm-c-test: 33.8384
clang-import-test: 20.6708
c-index-test: 19.8632
verify-uselistorder: 2.84348

vedant

On Mon, May 14, 2018 at 2:27 PM, Vedant Kumar via llvm-dev
<llvm-dev at lists.llvm.org> wrote:
> Should the binaries included in an LLVM release be generally useful to a
wide audience of toolchain users, or is it OK to ship tools that are only really
useful for LLVM development?
>
> If the former is the case, can we consider not shipping tools which are
exclusively for testing the compiler? E.g removing these 5 binaries from
5.0.2-rc1 would have saved over a hundred megabytes in uncompressed space for
macOS users:
>
> Binary name | Size in megabytes
> --------------------------------------------
> clang-check: 54.6725
> llvm-c-test: 33.8384
> clang-import-test: 20.6708
> c-index-test: 19.8632
> verify-uselistorder: 2.84348
>
> vedant
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
I personally have no objections to removing the tools you just pointed
out, as long as important testing tools are kept around (in
particular, opt, llc).

Thanks,

-- 
Davide

"There are no solved problems; there are only problems that are more
or less solved" -- Henri Poincare

Please keep llvm-symbolizer in the install distribution.

If a user's program crashes then I like to print out a backtrace with file
name/line number if compiled with debug. Forking off llvm-symbolizer is the
easiest way for me to obtain the debug information.

Thanks.


On Mon, May 14, 2018 at 5:27 PM, Vedant Kumar via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
> Should the binaries included in an LLVM release be generally useful to a
> wide audience of toolchain users, or is it OK to ship tools that are only
> really useful for LLVM development?
>
> If the former is the case, can we consider not shipping tools which are
> exclusively for testing the compiler? E.g removing these 5 binaries from
> 5.0.2-rc1 would have saved over a hundred megabytes in uncompressed space
> for macOS users:
>
> Binary name | Size in megabytes
> --------------------------------------------
> clang-check: 54.6725
> llvm-c-test: 33.8384
> clang-import-test: 20.6708
> c-index-test: 19.8632
> verify-uselistorder: 2.84348
>
> vedant
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180514/fd30cf95/attachment.html>

A new one was just added recently - Clang's diagtool. So if you're
looking
at how to reduce install size, might want to double check that the
motivation for adding that is consistent with/not contradictory with your
goals/motivations for removing these.

(also since several of these are clang binaries/tools - maybe check with
the cfe-dev list too)

On Mon, May 14, 2018 at 2:27 PM Vedant Kumar via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
> Should the binaries included in an LLVM release be generally useful to a
> wide audience of toolchain users, or is it OK to ship tools that are only
> really useful for LLVM development?
>
> If the former is the case, can we consider not shipping tools which are
> exclusively for testing the compiler? E.g removing these 5 binaries from
> 5.0.2-rc1 would have saved over a hundred megabytes in uncompressed space
> for macOS users:
>
> Binary name | Size in megabytes
> --------------------------------------------
> clang-check: 54.6725
> llvm-c-test: 33.8384
> clang-import-test: 20.6708
> c-index-test: 19.8632
> verify-uselistorder: 2.84348
>
> vedant
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180515/59efc193/attachment.html>

> On May 15, 2018, at 7:40 AM, David Blaikie <dblaikie at gmail.com>
wrote:
> 
> A new one was just added recently - Clang's diagtool. So if you're
looking at how to reduce install size, might want to double check that the
motivation for adding that is consistent with/not contradictory with your
goals/motivations for removing these.
Yep, the motivation for including clang's diagtool seemed sound to me (i.e I
can imagine diagtool being useful to a wide audience).

Actually, that change is what motivated me to take a closer look at what exactly
makes it into our installs.
> 
> (also since several of these are clang binaries/tools - maybe check with
the cfe-dev list too)
+ cfe-dev

vedant
> 
> On Mon, May 14, 2018 at 2:27 PM Vedant Kumar via llvm-dev <llvm-dev at
lists.llvm.org <mailto:llvm-dev at lists.llvm.org>> wrote:
> Should the binaries included in an LLVM release be generally useful to a
wide audience of toolchain users, or is it OK to ship tools that are only really
useful for LLVM development?
> 
> If the former is the case, can we consider not shipping tools which are
exclusively for testing the compiler? E.g removing these 5 binaries from
5.0.2-rc1 would have saved over a hundred megabytes in uncompressed space for
macOS users:
> 
> Binary name | Size in megabytes
> --------------------------------------------
> clang-check: 54.6725
> llvm-c-test: 33.8384
> clang-import-test: 20.6708
> c-index-test: 19.8632
> verify-uselistorder: 2.84348
> 
> vedant
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org>
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
<http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180515/01d77118/attachment.html>

> On May 14, 2018, at 4:14 PM, David Jones <david.jones at metrics.ca>
wrote:
> 
> Please keep llvm-symbolizer in the install distribution.
> 
> If a user's program crashes then I like to print out a backtrace with
file name/line number if compiled with debug. Forking off llvm-symbolizer is the
easiest way for me to obtain the debug information.
Yep, at this point I'm only considering removing tools which are strictly
for compiler testing.

vedant
> 
> Thanks.
> 
> 
> On Mon, May 14, 2018 at 5:27 PM, Vedant Kumar via llvm-dev <llvm-dev at
lists.llvm.org <mailto:llvm-dev at lists.llvm.org>> wrote:
> Should the binaries included in an LLVM release be generally useful to a
wide audience of toolchain users, or is it OK to ship tools that are only really
useful for LLVM development?
> 
> If the former is the case, can we consider not shipping tools which are
exclusively for testing the compiler? E.g removing these 5 binaries from
5.0.2-rc1 would have saved over a hundred megabytes in uncompressed space for
macOS users:
> 
> Binary name | Size in megabytes
> --------------------------------------------
> clang-check: 54.6725
> llvm-c-test: 33.8384
> clang-import-test: 20.6708
> c-index-test: 19.8632
> verify-uselistorder: 2.84348
> 
> vedant
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org>
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
<http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev>
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180515/60e8d124/attachment.html>

I'd like to argue in the other direction for the windows distribution, basic
tools like llvm-symbolizer are missing there, and it would be great if it was
included.

--Ray


On 5/14/2018 3:27 PM, Vedant Kumar via llvm-dev wrote:
> Should the binaries included in an LLVM release be generally useful to a
wide audience of toolchain users, or is it OK to ship tools that are only really
useful for LLVM development?
>
> If the former is the case, can we consider not shipping tools which are
exclusively for testing the compiler? E.g removing these 5 binaries from
5.0.2-rc1 would have saved over a hundred megabytes in uncompressed space for
macOS users:
>
> Binary name | Size in megabytes
> --------------------------------------------
> clang-check: 54.6725
> llvm-c-test: 33.8384
> clang-import-test: 20.6708
> c-index-test: 19.8632
> verify-uselistorder: 2.84348
>
> vedant
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev