PSA: clobber build needed for IDR_INSPECTOR_* compile errors in blink_platform_impl.cc

[Thiago Farina]

Hi chromiumers,

If you encountered an error like the following:

../../content/child/blink_platform_impl.cc:759:34: error: use of undeclared identifier 'IDR_INSPECTOR_OVERLAY_PAGE_HTML'
  { "InspectorOverlayPage.html", IDR_INSPECTOR_OVERLAY_PAGE_HTML,

you were affected by a CL of mine (crrev.com/285524).

The fix for it is simple, just do:

$ rm out/Debug/gen/webkit/grit/blink_resources.h

And it should get you back to the building track again.

Apologize for the unexpected breakage.

You can read more details about this in this thread:

https://groups.google.com/a/chromium.org/d/topic/chromium-dev/RLbePQp136s/discussion

Regards,

--

Thiago Farina

[tomm...@chromium.org]

Thanks for the PSA. I was (very briefly) confused.

[Daniel Cheng]

Out of curiosity, how come this worked on some bots but not others? I think this change is causing some of my tryjobs to fail: http://build.chromium.org/p/tryserver.blink/builders/mac_blink_compile_dbg/builds/14144/steps/compile/logs/stdio

On a related topic, can we make the build system handle these sorts of cases better? Requiring manual clobbers/cleanups of generated files is a pain. Landmines help a bit for keeping the bots green, but they aren't perfect--especially if you didn't know your change would have a breaking effect like this.

Daniel

--
--
Chromium Developers mailing list: chromi...@chromium.org
View archives, change email options, or unsubscribe:
http://groups.google.com/a/chromium.org/group/chromium-dev

[James Cook]

This is still affecting tryservers. If you see the compile error on your tryjobs you can force a clobber with "--clobber":

git cl try --bot=linux_chromium_dbg -m tryserver.chromium.linux --clobber

Jaes

To unsubscribe from this group and stop receiving emails from it, send an email to chromium-dev...@chromium.org.

[Thiago Farina]

On Tue, Jul 29, 2014 at 2:46 PM, James Cook <jame...@chromium.org> wrote:

This is still affecting tryservers. If you see the compile error on your tryjobs you can force a clobber with "--clobber":

I have sent an e-mail to chrome-troppers asking for a clobber.

If the landmines help, I will try to send a CL for it later tonight.

--

Thiago Farina

[James Robinson]

This was still happening as of 3 hours ago:

http://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_compile_dbg/builds/1131

--

--
Chromium Developers mailing list: chromi...@chromium.org
View archives, change email options, or unsubscribe:
http://groups.google.com/a/chromium.org/group/chromium-dev

[Eric Seidel]

Outages like this deserve a post-mortem.

A much better solution for this type of failure is to immediately roll out the affecting change and then only land it again when the build system has been taught to cope.  In this situation it sounds like we could have added a script for gyp to call as part of the build to just delete the header from the old location.

--

[Thiago Farina]

On Wednesday, July 30, 2014, Eric Seidel <ese...@chromium.org> wrote:

Outages like this deserve a post-mortem.

A much better solution for this type of failure is to immediately roll out the affecting change and then only land it again when the build system has been taught to cope.  In this situation it sounds like we could have added a script for gyp to call as part of the build to just delete the header from the old location.

They are cleaning the bots in crbug.com/398679. I agree a script would have been better.

--

Thiago Farina

[TAMURA, Kent]

Can we include blink_resources.h relative to gen/? Like

    #include "blink/public/resources/grit/blink_resources.h"

instead of

    #include "grit/blink_resources.h"

Why do we have multiple include paths for generated files?

--

--
Chromium Developers mailing list: chromi...@chromium.org
View archives, change email options, or unsubscribe:
http://groups.google.com/a/chromium.org/group/chromium-dev

--
TAMURA Kent
Software Engineer, Google

[Nico Weber]

On Wed, Jul 30, 2014 at 5:33 PM, TAMURA, Kent <tk...@chromium.org> wrote:

Can we include blink_resources.h relative to gen/? Like

    #include "blink/public/resources/grit/blink_resources.h"

instead of

    #include "grit/blink_resources.h"

Why do we have multiple include paths for generated files?

This sounds like a good idea to me. Let's see how it goes:

https://codereview.chromium.org/431883002

https://codereview.chromium.org/430083002

[Thiago Farina]

On Wed, Jul 30, 2014 at 9:33 PM, TAMURA, Kent <tk...@chromium.org> wrote:

Can we include blink_resources.h relative to gen/? Like

    #include "blink/public/resources/grit/blink_resources.h"

instead of

    #include "grit/blink_resources.h"

I had this idea, but haven't talk to anyone about it until you mentioned it.

I was thinking on it but for other reasons that Nico is doing it now. I had thought on switching to full path because it would make it easier for anyone to discover from where this header comes from. When you see grit/generated_resources.h you can't say it is from https://chromium.googlesource.com/chromium/src/+/master/chrome/app/generated_resources.grd, while if we had used full paths we could.

But of course doing this change will have some complications as the ones Nico is facing now.

--

Thiago Farina

[Dominic Mazzoni]

This is still affecting try servers:

http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_chromeos_dbg/builds/31/steps/compile%20%28with%20patch%29/logs/stdio#error1

[Nico Weber]

I just got this on a try run too, so I'm trying to add a landmine to fix this issue ONCE AND FOR ALL. If someone who knows how landmines work can review, that'd be nice: https://codereview.chromium.org/439503005/

On Fri, Aug 1, 2014 at 10:18 AM, Dominic Mazzoni <dmaz...@chromium.org> wrote:

This is still affecting try servers:

http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_chromeos_dbg/builds/31/steps/compile%20%28with%20patch%29/logs/stdio#error1

[Thiago Farina]

For those following this thread, it is possible that https://codereview.chromium.org/423613003/ might cause other issue like this. I'm not 100% it will, but I thought I should alert first.

--

Thiago Farina

[Nico Weber]

Don't land it without an action that makes it clean up after itself then? (Or wait until grit uses qualified includes before landing.)

[Eric Seidel]

This is happening again, this time on try-jobs:

crbug.com/400860

This needs-clobber nonsense has got to stop.  We need to fix the underlying build system problem.

[Julie Parent]

As Eric said earlier, this sort of outage benefits from a postmortem, so we can better understand the underlying problem and have explicit tasks to follow-up on, to prevent it from happening again.

Would someone involved in this outage like to guinea pig our new postmortem site?  It provides a template to start from, and a place to store your postmortem once it is complete: https://sites.google.com/a/chromium.org/dev/developers/postmortems

Julie

[Nico Weber]

And this time, if you're getting lots of

../../content/child/blink_platform_impl.cc:757:17: error: use of undeclared identifier 'IDR_UASTYLE_HTML_CSS'

  { "html.css", IDR_UASTYLE_HTML_CSS, ui::SCALE_FACTOR_NONE },

                ^

../../content/child/blink_platform_impl.cc:758:19: error: use of undeclared identifier 'IDR_UASTYLE_QUIRKS_CSS'

  { "quirks.css", IDR_UASTYLE_QUIRKS_CSS, ui::SCALE_FACTOR_NONE },

                  ^

../../content/child/blink_platform_impl.cc:759:24: error: use of undeclared identifier 'IDR_UASTYLE_VIEW_SOURCE_CSS'

  { "view-source.css", IDR_UASTYLE_VIEW_SOURCE_CSS, ui::SCALE_FACTOR_NONE },

                       ^

../../content/child/blink_platform_impl.cc:760:26: error: use of undeclared identifier 'IDR_UASTYLE_THEME_CHROMIUM_CSS'

  { "themeChromium.css", IDR_UASTYLE_THEME_CHROMIUM_CSS,

                         ^

the file to remove is out/{Release,Debug}/gen/blink/public/resources/grit/blink_resources.h

[Eric Seidel]

Dit the get_landmines change I landed not fix this for local builds?
https://codereview.chromium.org/441133002

[Nico Weber]

No, landmines are only used on bots.

[James Cook]

I saw a Mac builder hit the UASTYLE error about 3 hours ago -- should the landmine have handled that one?

http://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_compile_dbg/builds/3327

James

To unsubscribe from this group and stop receiving emails from it, send an email to chromium-dev...@chromium.org.

[Avi Drissman]

mac_chromium_compile_dbg is pretty broken. Failed again for me:

http://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_compile_dbg/builds/3479/steps/compile%20%28with%20patch%29/logs/stdio

[Christian Biesinger]

On Wed, Aug 6, 2014 at 4:49 PM, Nico Weber <tha...@chromium.org> wrote:
> And this time, if you're getting lots of
>
> ../../content/child/blink_platform_impl.cc:757:17: error: use of undeclared
> identifier 'IDR_UASTYLE_HTML_CSS'
> { "html.css", IDR_UASTYLE_HTML_CSS, ui::SCALE_FACTOR_NONE },
> ^
> ../../content/child/blink_platform_impl.cc:758:19: error: use of undeclared
> identifier 'IDR_UASTYLE_QUIRKS_CSS'
> { "quirks.css", IDR_UASTYLE_QUIRKS_CSS, ui::SCALE_FACTOR_NONE },
> ^
> ../../content/child/blink_platform_impl.cc:759:24: error: use of undeclared
> identifier 'IDR_UASTYLE_VIEW_SOURCE_CSS'
> { "view-source.css", IDR_UASTYLE_VIEW_SOURCE_CSS, ui::SCALE_FACTOR_NONE },
> ^
> ../../content/child/blink_platform_impl.cc:760:26: error: use of undeclared
> identifier 'IDR_UASTYLE_THEME_CHROMIUM_CSS'
> { "themeChromium.css", IDR_UASTYLE_THEME_CHROMIUM_CSS,
> ^
>
> the file to remove is
> out/{Release,Debug}/gen/blink/public/resources/grit/blink_resources.h

So... what should I do if removing
out/Debug/gen/blink/public/resources/grit/blink_resources.h did not
solve my build error?

-christian

[Thiago Farina]

On Wed, Aug 6, 2014 at 11:47 PM, Christian Biesinger <cbies...@google.com> wrote:

On Wed, Aug 6, 2014 at 4:49 PM, Nico Weber <tha...@chromium.org> wrote:
> And this time, if you're getting lots of
>
> ../../content/child/blink_platform_impl.cc:757:17: error: use of undeclared
> identifier 'IDR_UASTYLE_HTML_CSS'
>   { "html.css", IDR_UASTYLE_HTML_CSS, ui::SCALE_FACTOR_NONE },
>                 ^
> ../../content/child/blink_platform_impl.cc:758:19: error: use of undeclared
> identifier 'IDR_UASTYLE_QUIRKS_CSS'
>   { "quirks.css", IDR_UASTYLE_QUIRKS_CSS, ui::SCALE_FACTOR_NONE },
>                   ^
> ../../content/child/blink_platform_impl.cc:759:24: error: use of undeclared
> identifier 'IDR_UASTYLE_VIEW_SOURCE_CSS'
>   { "view-source.css", IDR_UASTYLE_VIEW_SOURCE_CSS, ui::SCALE_FACTOR_NONE },
>                        ^
> ../../content/child/blink_platform_impl.cc:760:26: error: use of undeclared
> identifier 'IDR_UASTYLE_THEME_CHROMIUM_CSS'
>   { "themeChromium.css", IDR_UASTYLE_THEME_CHROMIUM_CSS,
>                          ^
>
> the file to remove is
> out/{Release,Debug}/gen/blink/public/resources/grit/blink_resources.h

So... what should I do if removing
out/Debug/gen/blink/public/resources/grit/blink_resources.h did not
solve my build error?

I don't think you should remove that header file.

https://code.google.com/p/chromium/codesearch#chromium/src/out/Debug/gen/blink/public/resources/grit/blink_resources.h&q=IDR_UASTYLE_HTML_CSS&sq=package:chromium&l=5

-- 

Thiago Farina

[Christian Biesinger]

Oh well, I solved my personal problem with rm -rf out.

-christian

[Jochen Eisinger]

I filed crbug.com/400011 about developer checkouts respecting landmines.

Maybe this should be part of the CY?

best

-jochen

[Julie Parent]

crbug.com/400011 would fall under Ojan's "Tree always open" part of the CY, as it would help prevent the tree from being closed like this in the future.

The other aspect of this that DOES clearly fall into the Code Yellow is the need for a postmortem.  It sounds like we understand what went wrong here, so we just need someone who was involved (tfarina?) to synthesize the threads and related follow-up bug(s) into a clear postmortem.  This isn't about shaming anyone, but rather, it is a learning experience for the whole project.  We'd like to have postmortems generated regularly (well, we'd rather not have outages at all ... but when they do, we want postmortems to better understand what we wrong, and put us in a place with clear followups so that they don't go wrong in the same, or similar, ways again).   You can start with the template on the new postmortem site (https://sites.google.com/a/chromium.org/dev/developers/postmortems), and once it is complete, please add it to the list on that page, and send it out to chromium-dev.

[Thiago Farina]

On Thu, Aug 7, 2014 at 6:06 PM, Julie Parent <jpa...@chromium.org> wrote:

 You can start with the template on the new postmortem site (https://sites.google.com/a/chromium.org/dev/developers/postmortems), and once it is complete, please add it to the list on that page, and send it out to chromium-dev.

Heads up: We are working on this. Will post here on chromium-dev when it is ready.

-- 

Thiago Farina