When first introduced, test-verify was only run when .js/.html/.xul/.xhtml files
were changed. Recently, it seems to run on every push. This is a speculative fix:
There may be confusion between "test-verify" and "test-verification" so I am
using "test-verify" consistently.
By default, run_process() eats the output of the child process,
which means when it fails we get an exception about the return
code, but none of the error messages the script printed, with
makes the cause of failure opaque.
Resolve this by passing a log_name parameter, which will cause
the default line_handler to pass all lines of output to the
local logging instance, by default at the INFO level.
MozReview-Commit-ID: FIVRggeKT4f
--HG--
extra : rebase_source : 813f578aeb1801a9508eb0cb1cd5e9ca1429148e
Bug 1365460 introduced code paths behind MOZ_DEBUG #ifdefs, but
MOZ_DEBUG is never defined, while it is available in CONFIG in
moz.builds. This is kind of a confusing situation, but the fact that
we've been able to avoid those problems for so long would tend to
put the blame on mozjemalloc, and fixes should go there.
Except that bug 1261161 explains that the only existing alternative
(the DEBUG #define), as used in MFBT, is not working for spidermonkey,
so it actually makes sense to converge to MOZ_DEBUG rather than DEBUG.
So start defining MOZ_DEBUG globally, fixing the mozjemalloc issues of
not having the debug code enabled. Bug 1261161 can then take care of
changing the DEBUG #ifdefs.
--HG--
extra : rebase_source : 37e3d03ac8350c62c8059d4ca01d1ecfdf5f421a
The feature should be pretty self-explanatory. It will enable easier
machine readable output.
MozReview-Commit-ID: GK148VCcNjm
--HG--
extra : rebase_source : ae5b241f275cfe3038ec202c080796270465c770
I likely regressed this as part of e9416a307987 (bug 1397406).
Before this change, '*' in filenames didn't work.
MozReview-Commit-ID: 33m83H6UTZ
--HG--
extra : rebase_source : 5732cd2b7f4851c6cd564c8dcbd7d303fbad570b
This fixes a regression from bug 1408365. Ideally these imports would
be defined at the top of the file, then the py2 linter would have caught
them. Even more ideally, mozboot would have some tests that hit these
code paths.
MozReview-Commit-ID: BWvIwAdUBF4
--HG--
extra : rebase_source : 6472f730a5cd12aa98af7a21f958d1ad2400f995
I think we're approaching an inflection point for the bootstrapper,
one where it no longer is possible to bootstrap without a source
checkout. For now, however, let's just do the simplest thing and
install the Proguard JAR along the happy path.
MozReview-Commit-ID: xUY37eE6oR
--HG--
extra : rebase_source : 31549ab3b6d662d84761c2a260cd236a5809c8ac
The goal is to use a newer Android-Gradle build plugin version (2.3.3
is latest stable). That requires a modern Gradle (anything 3.3+, but
3.4.1 is the default from my Android Studio), and also a newer
build-tools (25.0.3 is latest stable).
The locations of lint output changed, and we want to use the standard
output location because it's difficult to accommodate variant details
in custom names. We change the location of findbugs output to follow
suit.
This requires either:
- fixing lint errors
- adding to the lint whitelist
- using the new lint baseline
It's best to use the new lint baseline, which will happen in the next commit.
MozReview-Commit-ID: D19FzIDCJrE
--HG--
extra : rebase_source : 12d132c0c3e0dbe2b8873b31360ea96d612de44c
By default Eclipse CDT will scan the source tree for binaries so that it can
add those binaries to the list of things that it can run. This scanning is a
constant interuption and can last several seconds. Besides that, it's
currently useless for our setup since the only binaries that we're interested
in running are in the object directory (which it doesn't scan), and those are
set up during project generation. (The only binaries found in the source tree
are a couple of uninteresting bundled libraries.)
CLOSED TREE
MozReview-Commit-ID: 2WaH8qceALq
By default, scalability mode is activated for files with 5,000 lines or more.
There are quite a few C++ files with more than 5,000 lines, and Eclipse seems
to work fine with them with scalability mode turned off (even
nsCSSFrameConstructor.cpp which is over 13,000 lines long). Increasing the
number of lines before scalability mode is enabled allows Eclipse to handle
these files better.
MozReview-Commit-ID: 8mGYIHStHes
Without this setting, eclipse will refuse to open Objective-C files (it will
defer to an external editor). Adding *.mm to the file types that are treated
as C++ allows Eclipse to open them, and to provide some code assistance for
the bits of the files that it can understand.
MozReview-Commit-ID: ASeXesWxY4g
This setting allows Eclipse to notice when files it has open are changed
externally (such as by hg/git, for example). It can then update the contents
that it has for the open files, avoiding annoying issues such as saving changes
after an `hg pull -u` resulting in either "Resource is out of sync" errors or
else clobbering of the changes hg made to files.
MozReview-Commit-ID: 8WmewM7lbHe
The blocking Welcome screen is quite confusing for a new user. It is not clear
where to find the Mozilla stuff they expect to see when opening Eclipse, or
that the user needs to close the entire content area to get to it. Besides that
the screen isn't very useful for Mozilla people who will find more relevant
help from searching the online documentation, and who won't be creating new
projects in the generated workspace, etc.
MozReview-Commit-ID: 8YssnonAR1d
The setting to ensure that there is a newline at the end of files when they
are saved is very annoying. Besides adding unnecessary and unexpected cruft
to diffs, some parts of the codebase (some of the reftest directories for
example) have a policy of NOT having a newline at the end of their files.
MozReview-Commit-ID: IjIYxDsKS13
This makes us write the code formatter settings into the workspace settings
directory instead of the project settings directory. This is preferable
since when users make settings changes they are more likely to work with the
workspace settings, so we should put them there. Putting them there also
fixes a bug whereby the calls to _write_noindex/_remove_noindex would
overwrite the formatter settings file shortly after it had been created.
To get the formatter to show up in the UI we also need to set the formatter
settings as a one line pref value in the CDT UI settings. This duplication
is what Eclipse does when a new formatter is manually added, and it's
necessary to get the formatter working correctly.
MozReview-Commit-ID: KP4w0VbNCF7
We expect to start requiring rust 1.21 soon, so have
./mach bootstrap upgrade users to it to consolidate
churn between now and then.
MozReview-Commit-ID: HEnXm25duUr
--HG--
extra : rebase_source : d05071ee5c2852eb69da22e80623f21ca8c6f7cb
For some reason, Windows builders are setup in a way that prevents a
task from purging the toolchain cache of old files. Until that is
figured out and fixed, all the error message related to that achieves is
confuse people because the treeherder Failure Classification tab shows
them first. Practically speaking, the error doesn't prevent anything
from working properly. The worst that can happen is disk space running
out because of the toolchain cache being filled up, which would lead to
actual errors.
As the error happens on very many windows build, that leads to a lot of
errors being bucketed on bug 1391811, while they may well be varying
unrelated issues.
It also leads to people thinking their try builds fail because of that,
when they aren't (e.g. bug 1408212).
--HG--
extra : rebase_source : e25fc99db8e0920cfa238d0f78f15c78e140e3ec
The old system was simply in place because I couldn't figure out how
to pipe `yes | ...` in Python. This is good enough to replace it, and
much less fragile since the license hashes change frequently.
MozReview-Commit-ID: AhJJPqMKfUh
--HG--
extra : rebase_source : 86289e692d646181d545457fc953610e165ee2df
This is just one flavor of the "reftets" suite, so we need to add a distinct
scheduling component for it.
MozReview-Commit-ID: AtKuvuUCk1l
--HG--
extra : rebase_source : 3f316f0293e8d1245fc6e891bbcd044586ab6c06
This was simply oversight before. I ran into this using the
taskcluster-proxy /bewit interface, which returns a URL of the form
https://domain.net/short/path/to.file?bewit="several thousand
characters", which leads to an IOError due to the long path. Let's
assume that such query strings and fragments are transient; we should
drop these parts of the fetched URLs when writing to disk.
MozReview-Commit-ID: FMJHMp7a3rA
--HG--
extra : rebase_source : da701e7d5250f59f39e4b6262952a08d0068e9ac