Note that the GN and the GN build were not exactly from the same revision so some binaries might have been added between them, but I know that at least some of these dependencies are invalid (like bro.exe).

Was this a component build? If so it is known and expected that many extra DLLs get added. See  crbug.com/586967  and this line of code:

https://cs.chromium.org/chromium/src/chrome/tools/build/win/create_installer_archive.py?q=f:create_install&sq=package:chromium&l=513

I don't know why bro.exe is getting built, but if it is now being used to compress some assets then that may be by design.

Never mind.  crbug.com/586967  causes DLLs to be pulled in if they exist, but it never causes them to be built.

For the record, the reason why I've noticed this is because I was hitting a build warning in bro.exe (fixed in https://codereview.chromium.org/2027873006/) while making a GN build of mini_installer (args="is_official_build=true is_debug=false") and szabadka@ told me that there was no reason to build bro.exe with Chrome. I also don't think that a typical build of Chrome needs mojo_runner ? 

I did clean builds (rmdir /s/q out\Debug and gn clean out\Debug_gn_component) of component debug builds with gyp and gn and compared the DLLs and EXEs created. The gn builds generate 15 more DLLs and 1 more executable. The actual comparison is messier because the gn builds aren't a strict superset (renaming of outputs? Other reasons?) - the lists are attached. For executables the differences are:

gyp creates nacl64.exe and wow_helper.exe that gn doesn't
gn creates frame_analyzer.exe, mojo_runner.exe, and rgba_to_i420_converter.exe that gyp doesn't

As of 903b1d2f36 they both create bro.exe.

Deleted: dlls_gn.txt 2.4 KB

dlls_gn.txt 2.4 KB View Download

Deleted: dlls_gyp.txt 2.2 KB

dlls_gyp.txt 2.2 KB View Download

Deleted: exes_gn.txt 282 bytes

exes_gn.txt 282 bytes View Download

Deleted: exes_gyp.txt 247 bytes

exes_gyp.txt 247 bytes View Download

Thanks for looking at this ! Should we open a different bug for each of these different executables ?

Probably best to keep it as one bug - the causes are likely to be related.

If the actual 'correct' set of dependencies is the intersection of the gn/gyp results then we can potentially make everything faster than before which would be great.

I'm guessing this doesn't need to block shipping GN; do we need to fix this before we can drop GYP?

Nop, it doesn't prevent us from shipping, we should just look at this one day to improve the build time.

This issue has been Available for over a year. If it's no longer important or seems unlikely to be fixed, please consider closing it out. If it is important, please re-triage the issue.

Sorry for the inconvenience if the bug really should have been left as Available. If you change it back, also remove the "Hotlist-Recharge-Cold" label.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

A lot has changed since this bug was filed. bro.exe is now called brotli.exe. The other three executables are not created on a build of chrome.exe, but test_ime_driver.service.exe is built.

This command gives some hints as to why brotli.exe is built:

>gn path . //third_party/brotli:brotli //chrome:chrome_initial
//chrome:chrome_initial --[public]-->
//chrome:packed_resources --[public]-->
//chrome:packed_resources_extra --[private]-->
//content/browser/devtools:devtools_resources --[public]-->
//content/browser/devtools:devtools_resources_grit --[private]-->
//content/browser/devtools:compressed_protocol_json --[private]-->
//third_party/brotli:brotli

So, I guess we are compressing some .json files, which is good.

I'm not sure why test_ime_driver.service.exe is being built. What am I doing wrong here?

>gn path . //services/ui/ime/test_ime_driver:test_ime_driver //chrome:chrome_initial
No non-data paths found between these two targets.

The only one of the DLLs that were originally listed that I see being built is osmesa.dll. Note that I was building chrome.exe rather than mini_installer. I was doing a debug non-component build.