Make sure LNT does not fail when test result is empty.

llvm-svn: 213128

LNT shows geomean for all tests, even if user filters a subset of tests.

This patch lets LNT to...

LNT ignores small changes by default to save users from getting
excessive amount of results. How...

llvm-svn: 212985

llvm-svn: 212984

llvm-svn: 212746

LNT client was reporting everything as UNCHANGED_PASS, because a
comparison_window of 0, each re...

Both multisample and LNT report need ways to update reports. This patch
adds some simple functi...

llvm-svn: 212739

llvm-svn: 212738

llvm-svn: 212737

Rename results to server_report, to help distinguish between that and the local report object.

...

Update comments to reflect changes to geomean calculation.

No functionality change intended.

l...

llvm-svn: 212295

Baseline plot is currently quite confusing. Second baseline and those after
has color irrelevant...

- We now log a more readable diagnostic in situations that error during normal
invocation of...

llvm-svn: 211706

llvm-svn: 211603

It is not useful to report failure on empty sample or when not comparing to
other runs.

llvm-sv...

llvm-svn: 211582

Return samples as an empty list when geomean is None.

llvm-svn: 211580

llvm-svn: 211148

not_none was not correctly implemented.

llvm-svn: 210352

This patch enables LNT to use benchmarking only mode in test suite.

llvm-svn: 210276

llvm-svn: 210139

llvm-svn: 210136

llvm-svn: 210134

Geomean calculation was having problems with compile test suite. Fix that, as well as some bugs...

llvm-svn: 210131

llvm-svn: 210128

Geomean calculation was having problems with compile test suite. Fix that, as well as some bugs...

llvm-svn: 210122

Fix the regression from r209204, removes no longer used comparison_window.

llvm-svn: 210088

In the event of a baseline request which produces no data, don't show that baseline.

llvm-svn: ...

This method significantly reduces noise when applied in combination of
Mann-Whitney U test. It r...

llvm-svn: 209118

We are using jQuery everywhere else, rewrite using jQuery for consistency.

llvm-svn: 209071

llvm-svn: 209041

Use of jquery-latest is not recommended
Replace the use of 1.5 with 1.7 to fix some rendering is...

llvm-svn: 209021

llvm-svn: 209020

llvm-svn: 209019

llvm-svn: 209018

llvm-svn: 208738

Turns out we have a Mann-Whitney U test implementation(for large sample
size) in external.stats....

Background colour of pct_delta was not rendered on odd numbered lines,
because HTML bgcolor tag ...

0.05 confidence level is too strict on noisy machines, add 0.10 as an extra
option.

llvm-svn: 2...

Mouse wheel zoom is disabled, so remove the instructions for it.

llvm-svn: 208612

SciPy uses one-sided p-value while rest of program uses two-sided, multiplied
by 2.

llvm-svn: 2...

Forgot to update a variable during code refactoring

llvm-svn: 208534

Correctly calculates Mann-Whitney U
Optionally depends on SciPy

llvm-svn: 208450

Flot's mouse wheel zoom feature does not work well on OSX. Since there is a manual zoom button, ...

The mouse wheel zooming can be cumbersome on some platforms. This change provides good ol' fashi...

When there are many baselines on a LNT graph, they become hard to distinguish. This changes base...

When multiple baseline are present, they are hard to distinguish from each other. This change a...

llvm-svn: 208125

llvm-svn: 207918

llvm-svn: 207898

llvm-svn: 207773

The command "lnt runtest compile" crashes, where as "lnt runtest nt" prints help. Fix this so th...

Add visual representation of baselines in graphs. When passing a "baseline.<n>=<run_num>" argum...

To help diagnose server failures more easily.

llvm-svn: 204367

If passing json=true to the graphing page, export the graph data as json instead of HTML.

llvm-...

Adds a --qemu-user-mode=VALUE option that sets USER_EMULATION_MODE to '1' and
RUNUNDER to the gi...

In a previous refactoring responsibility for printing during submission was moved to the callers...

Lnt submit is used for sending results to the server, so having the default behavior be not subm...

This will allow for future validation of modified Report objects.

llvm-svn: 199270

Change submit to return results from the server, so they can be used to determine needed reruns....

llvm-svn: 198668

llvm-svn: 198667

With resampling the submission logic will change in NT.
Deprecate runtest command line args. Mov...

This reverts commit r195509. As was discussed on the list, the output from subcommands makes th...

This commit broke the following LNT buildbots:

clang-x86_64-darwin12-nt-O3
clang-native-arm-lnt...

This is an attempt to fix the build after commit r198467.

llvm-svn: 198483

llvm-svn: 198467

When we don't commit runs, don't pass back the run.id, because the run won't exist.

llvm-svn: 1...

llvm-svn: 198464

llvm-svn: 198463

llvm-svn: 198460

llvm-svn: 198458

llvm-svn: 198455

It is nice to know the runid on the client, for easy lookup if on the server later. This commit...

With resampling the submission logic will change in NT.
Deprecate runtest command line args. Mov...

This looks like just a careless mistake where the interpolation pattern was not
placed in correc...

llvm-svn: 196188

llvm-svn: 195531

llvm-svn: 195530

llvm-svn: 195527

llvm-svn: 195509

llvm-svn: 195506

llvm-svn: 195498

llvm-svn: 194518

llvm-svn: 194517

When you specified something odd for --cc you'd get a really cryptic "Permission Denied" error w...

llvm-svn: 192726

llvm-svn: 190020

llvm-svn: 190018

- This limits the set of test names we consider (and thus report on) to be
tests that appear...

llvm-svn: 189533

llvm-svn: 188399