11.07.2015 Views

Comparison of Seven Bug Report Types: A Case-Study of ... - IIIT

Comparison of Seven Bug Report Types: A Case-Study of ... - IIIT

Comparison of Seven Bug Report Types: A Case-Study of ... - IIIT

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

TABLE VIEXAMPLES OF BUG TYPES FROM GOOGLE CHROMIUM BROWSER PROJECT (POLISH, SECURITY, CLEANUP, REGRESSION)Type Issue ID Text Snippet from <strong>Bug</strong> <strong>Report</strong>Polish 94114 Several infobars are lacking punctuationPolish 51429 Bookmarks bubble (and extensions installed bubble) should have white backgroundPolish 89591 Misaligned text in HTML file input controlSecurity 1208 puts users at more risk from websites trying to spo<strong>of</strong> a file’s typeSecurity 23693 if the user is duped into clicking ”Create application shortcut” on a malicious pageSecurity 23979 We consider some file extensions to be maliciousSecurity 24733 because a given page can deterministically take down the whole browserSecurity 74665 This can lead to memory corruption and is probably exploitable too if you are clevercleanup 88098 Use javascript objects to make the ui code cleaner and easier to modifycleanup 30151 Toolstrip is obsolete, and these tests should be removed when toolstrip is removedcleanup 68882 src/gfx uses wstring all over. It should use string16Regression 115321 Tab crashes when coming out <strong>of</strong> the Full screen for the Sublime videosRegression 115934 Info bar <strong>of</strong>fers to save a wrong passwordTABLE VIIFREQUENT TERMS[CHROME-BROWSER]Top K CRASH REGR. SECUR. CLEAN POLISH PERF. USAB.1 crash regress secur should should regress page2 chrome crash crash remov bookmark page window3 report page chrome test button mac user4 stack tab us us window perf chrome5 signatur chrome memori code text time should6 webcor window corrupt move tab sync zoom7 browser bookmark file chrome menu test open8 const doesnt audit webkit bar releas doe9 Tab browser browser api dialog cycler manag10 Int open webcor need mac startup menu11 render work page browser page slow bookmark12 std render access content chrome tab provid13 mac broken window add need linux url14 page new webkit base icon new line15 ... text free clean drag veri dialog16 googl bar sandbox extens us chrome select17 intern show render refactor manag load command18 open button url view open us click19 char click bypass file select theme us20 unsign std open renam close cpu pluginTABLE VIIIDESCRIPTION LENGTHCRASH REGR. SECUR. CLEAN POLISH PERF. USAB.Min 7 5 22 3 11 54 64Q1 556 350 328 117.5 234.75 304.25 404Median 1021 550 743 225 396.5 537.5 593Q3 2909 859 1436 422.5 603 899.5 876Max 20057 19875 16233 9551 1837 3719 2886the description length <strong>of</strong> cleanup bug reports is the lowest.<strong>Bug</strong> reports belonging to crash types mostly consists <strong>of</strong> systemgenerated crash reports and hence size <strong>of</strong> description is largeas compare to other bug types where users have to provideentire content manually.D. Correlation Between Regression and other <strong>Bug</strong> <strong>Types</strong>A regression bug (one <strong>of</strong> the seven types <strong>of</strong> bug reportsin the evaluation dataset) can also belong to other categories(such as crash, polish and performance). We conduct experimentsto identify relation between regression bugs and othertypes <strong>of</strong> bugs. The motivation behind the experiment is toidentify bug-types which are more prone to cause regressionbugs. We extract all bug-type labels present in regression bugreports and then compute correlations. Table IX shows results<strong>of</strong> our experiment. Table IX reveals relatively high correlationbetween regression bugs and crash bugs (nearly 7% <strong>of</strong> all theregression bugs are crash bugs), however manual observation<strong>of</strong> 100 bugs (fixed/verified) shows different correlation. Wediscover that out <strong>of</strong> 100 regression bugs only 16 bugs belongto crash type and 43 bugs belong to the usability type. Furthermore,there are 31 bugs which can be categorized to any<strong>of</strong> usability, polish, or cleanup (based on our interpretation <strong>of</strong>bug description/discussion). We believe that a large number523

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!