914925 - mach reftest --debugger doesn't work

Status: RESOLVED FIXED

(Testing :: Reftest, defect)

Description

[Cameron McCormack (:heycam)]

Same as bug 913173, but for reftest.

master irk:/z/moz/central $ ./mach reftest --debugger=gdb
 0:00.20 /usr/bin/make -C . -j8 -s -w reftest
make: Entering directory `/z/moz/central/obj-mac-dbg'
REFTEST INFO | runreftest.py | Running tests: start.

INFO | automation.py | Launching: /usr/bin/gdb -q --args /z/moz/central/obj-mac-dbg/dist/NightlyDebug.app/Contents/MacOS/firefox-bin -foreground -no-remote -profile /var/folders/04/rtvfhstx5c1_y4sv_179m40c0000gn/T/tmpFHBM2J.mozrunner/ -reftest /z/moz/central/layout/reftests/reftest.list
INFO | automation.py | Application pid: 20356
TEST-INFO: Not logging stdout or stderr due to debugger connection
Reading symbols for shared libraries ......... done
(gdb) r
Starting program: /z/moz/central/obj-mac-dbg/dist/NightlyDebug.app/Contents/MacOS/firefox-bin -foreground -no-remote -profile /var/folders/04/rtvfhstx5c1_y4sv_179m40c0000gn/T/tmpFHBM2J.mozrunner/ -reftest /z/moz/central/layout/reftests/reftest.list
...
Program exited with code 06.
(gdb)

Comment 1

[Cameron McCormack (:heycam)]

I tried sticking in something like bug 913173 comment 11 in runreftest.py's buildBrowserEnv but this didn't have any effect.

Comment 2

[(not currently active) Ted Mielczarek]

Attachment: fix --debugger for reftest/mochitest on OS X

I got motivated and did a Mac build and could reproduce the problem (in both reftest and mochitest). This patch fixes the problem.

Comment 3

[Joel Maher ( :jmaher ) (UTC -8)]

Comment on attachment 803185 [details] [diff] [review]
fix --debugger for reftest/mochitest on OS X

Review of attachment 803185 [details] [diff] [review]:
-----------------------------------------------------------------

::: layout/tools/reftest/runreftest.py
@@ +107,5 @@
>      )
>      self.copyExtraFilesToProfile(options, profile)
>      return profile
>  
> +  def buildBrowserEnv(self, options, profileDir, debugger):

why are you not setting this to a default value of False as you do in the mochitest runner?

Comment 4

[(not currently active) Ted Mielczarek]

I didn't think it was necessary since this was the only caller, but I forgot that runtestsremote calls this as well. I'll fix that, thanks!

Comment 5

[(not currently active) Ted Mielczarek]

Oh, actually, I guess nothing does call the reftest version, but I'll fix it just for sanity.

Comment 6

[(not currently active) Ted Mielczarek]

Attachment: fix --debugger for reftest/mochitest on OS X.

Comment 7

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/integration/fx-team/rev/9157ea42ff41

Comment 8

[Ryan VanderMeulen [:RyanVM]]

Backed out for remote test bustage.
https://hg.mozilla.org/integration/fx-team/rev/a23d608fb893

Comment 10

[Matt Woodrow (:mattwoodrow)]

This patch doesn't seem to fix the problem for me, still can't debug reftests.

Comment 11

[(not currently active) Ted Mielczarek]

You need to apply the patch then rebuild the build/ directory to get automation.py updated. I tested on my Mac and it worked, so I suspect it's just horrible preprocessed Python nonsense.

Comment 12

[Adam Roach [:abr]]

The patch seems to have rotted, and the naïve manual application of the reject files doesn't seem to produce a viable environment.

Comment 13

[(not currently active) Ted Mielczarek]

Attachment: fix --debugger for reftest/mochitest on OS X.

Rebased to work on top of the Mochitest refactor.

Comment 14

[Bobby Holley (:bholley)]

I still need to manually apply this patch to basically every repo I work in, FWIW.

Comment 15

[Bobby Holley (:bholley)]

At the summit, Ted promised to land something to make mochitests work with a debugger again.

Comment 16

[(not currently active) Ted Mielczarek]

Attachment: fix --debugger for mochitest on OS X.

This patch only touches Mochitest. This should be simpler to land.

Comment 17

[(not currently active) Ted Mielczarek]

*sigh*, that still broke Android tests:
https://tbpl.mozilla.org/?tree=Try&rev=2cd6ba574f2e

Comment 18

[Bobby Holley (:bholley)]

Can we get some traction here? Manually shuffling this patch around in local trees is a pain (especially because it borks android try pushes), but more importantly, this is mysteriously breaks the experience of any newcomer dev who doesn't know about the workaround.

Comment 19

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

I'm looking into fixing android

Comment 20

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

Attachment: Patch v2

Comment 21

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

https://tbpl.mozilla.org/?tree=Try&rev=d8a3663659bb

Comment 22

[Joel Maher ( :jmaher ) (UTC -8)]

Comment on attachment 8334493 [details] [diff] [review]
Patch v2

Review of attachment 8334493 [details] [diff] [review]:
-----------------------------------------------------------------

this looks to be the right patch.

Comment 23

[Boris Zbarsky [:bzbarsky]]

Ms2ger, thank you!

Comment 24

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/9c31c75b81ea

Comment 25

[Bobby Holley (:bholley)]

Ms2ger, you are a hero.

Comment 26

[(not currently active) Ted Mielczarek]

Thanks for that, I owe you one.

Comment 27

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/9c31c75b81ea