1 # Chrome Network Bug Triage : Suggested Workflow
5 ## Looking for new crashers
7 1. Go to [go/chromecrash](https://goto.google.com/chromecrash).
9 2. For each platform, look through the releases for which releases to
10 investigate. As per bug-triage.txt, this should be the most recent canary,
11 the previous canary (if the most recent is less than a day old), and any of
12 dev/beta/stable that were released in the last couple of days.
14 3. For each release, in the "Process Type" frame, click on "browser".
16 4. At the bottom of the "Magic Signature" frame, click "limit 1000". Reported
17 crashers are sorted in decreasing order of the number of reports for that
20 5. Search the page for *"net::"*.
22 6. For each found signature:
23 * If there is a bug already filed, make sure it is correctly describing the
24 current bug (e.g. not closed, or not describing a long-past issue), and
25 make sure that if it is a *net* bug, that it is labeled as such.
26 * Ignore signatures that only occur once, as memory corruption can easily
27 cause one-off failures when the sample size is large enough.
28 * Ignore signatures that only come from a single client ID, as individual
29 machine malware and breakage can also easily cause one-off failures.
30 * Click on the number of reports field to see details of crash. Ignore it
31 if it doesn't appear to be a network bug.
32 * Otherwise, file a new bug directly from chromecrash. Note that this may
33 result in filing bugs for low- and very-low- frequency crashes. That's
34 ok; the bug tracker is a better tool to figure out whether or not we put
35 resources into those crashes than a snap judgement when filing bugs.
36 * For each bug you file, include the following information:
37 * The backtrace. Note that the backtrace should not be added to the
38 bug if Restrict-View-Google isn't set on the bug as it may contain
39 PII. Filing the bug from the crash reporter should do this
40 automatically, but check.
41 * The channel in which the bug is seen (canary/dev/beta/stable), its
42 frequency in that channel, and its rank among crashers in the
44 * The frequency of this signature in recent releases. This information
46 1. Clicking on the signature in the "Magic Signature" list
47 2. Clicking "Edit" on the dremel query at the top of the page
48 3. Removing the "product.version='X.Y.Z.W' AND" string and clicking
50 4. Clicking "Limit 1000" in the Product Version list in the
51 resulting page (without this, the listing will be restricted to
52 the releases in which the signature is most common, which will
53 often not include the canary/dev release being investigated).
54 5. Choose some subset of that list, or all of it, to include in the
55 bug. Make sure to indicate if there is a defined point in the
56 past before which the signature is not present.
58 ## Identifying unlabeled network bugs on the tracker
60 * Look at new uncomfirmed bugs since noon PST on the last triager's rotation.
61 [Use this issue tracker
62 query](https://code.google.com/p/chromium/issues/list?can=2&q=status%3Aunconfirmed&sort=-id&num=1000).
64 * Press **h** to bring up a preview of the bug text.
66 * Use **j** and **k** to advance through bugs.
68 * If a bug looks like it might be network/download/safe-browsing related,
69 middle click (or command-click on OSX) to open in new tab.
71 * If a user provides a crash ID for a crasher for a bug that could be
72 net-related, look at the crash stack at
73 [go/crash](https://goto.google.com/crash), and see if it looks to be network
74 related. Be sure to check if other bug reports have that stack trace, and
75 mark as a dupe if so. Even if the bug isn't network related, paste the stack
76 trace in the bug, so no one else has to look up the crash stack from the ID.
77 * If there's no other information than the crash ID, ask for more details
78 and add the Needs-Feedback label.
80 * If network causes are possible, ask for a net-internals log (If it's not a
81 browser crash) and attach the most specific internals-network label that's
82 applicable. If there isn't an applicable narrower label, a clear owner for
83 the issue, or there are multiple possibilities, attach the internals-network
84 label and proceed with further investigation.
86 * If non-network causes also seem possible, attach those labels as well.
88 ## Investigating Cr-Internals-Network bugs
90 * It's recommended that while on triage duty, you subscribe to the
91 Cr-Internals-Network label. To do this, go to the issue tracker and click on
92 [Subscriptions](https://code.google.com/p/chromium/issues/subscriptions).
93 Enter "Cr-Internals-Network" and click submit.
95 * Look through uncomfirmed and untriaged Cr-Internals-Network bugs,
96 prioritizing those updated within the last week. [Use this issue tracker
97 query](https://code.google.com/p/chromium/issues/list?can=2&q=Cr%3DInternals-Network+-status%3AAssigned+-status%3AStarted+-status%3AAvailable+&sort=-modified).
99 * If more information is needed from the reporter, ask for it and add the
100 Needs-Feedback label. If the reporter has answered an earlier request for
101 information, remove that label.
103 * While investigating a new issue, change the status to Untriaged.
105 * If a bug is a potential security issue (Allows for code execution from remote
106 site, allows crossing security boundaries, unchecked array bounds, etc) mark
107 it Type-Bug-Security. If it has privacy implication (History, cookies
108 discoverable by an entity that shouldn't be able to do so, incognito state
109 being saved in memory or on disk beyond the lifetime of incognito tabs, etc),
112 * For bugs that already have a more specific network label, go ahead and remove
113 the Cr-Internals-Network label and move on.
115 * Try to figure out if it's really a network bug. See common non-network
116 labels section for description of common labels needed for issues incorrectly
117 tagged as Cr-Internals-Network.
119 * If it's not, attach appropriate labels and go no further.
121 * If it may be a network bug, attach additional possibly relevant labels if
122 any, and continue investigating. Once you either determine it's a
123 non-network bug, or figure out accurate more specific network labels, your
124 job is done, though you should still ask for a net-internals dump if it seems
127 * Note that ChromeOS-specific network-related code (Captive portal detection,
128 connectivity detection, login, etc) may not all have appropriate more
129 specific labels, but are not in areas handled by the network stack team.
130 Just make sure those have the OS-Chrome label, and any more specific labels
131 if applicable, and then move on.
133 * Gather data and investigate.
134 * Remember to add the Needs-Feedback label whenever waiting for the user to
135 respond with more information, and remove it when not waiting on the
137 * Try to reproduce locally. If you can, and it's a regression, use
138 src/tools/bisect-builds.py to figure out when it regressed.
139 * Ask more data from the user as needed (net-internals dumps, repro case,
140 crash ID from about:crashes, run tests, etc).
141 * If asking for an about:net-internals dump, provide this link:
142 https://sites.google.com/a/chromium.org/dev/for-testers/providing-network-details.
143 Can just grab the link from about:net-internals, as needed.
145 * Try to figure out what's going on, and which more specific network label is
148 * If it's a regression, browse through the git history of relevant files to try
149 and figure out when it regressed. CC authors / primary reviewers of any
150 strongly suspect CLs.
152 * If you are having trouble with an issue, particularly for help understanding
153 net-internals logs, email the public net-dev@chromium.org list for help
154 debugging. If it's a crasher, or for some other reason discussion needs to
155 be done in private, use chrome-network-debugging@google.com. TODO(mmenke):
156 Write up a net-internals tips and tricks docs.
158 * If it appears to be a bug in the unowned core of the network stack (i.e. no
159 sublabel applies, or only the Cr-Internals-Network-HTTP sublabel applies, and
160 there's no clear owner), try to figure out the exact cause.
162 ## Monitoring UMA histograms and gasper alerts
164 For each Gasper alert that fires, determine if it's a real alert and file a bug
167 * Don't file if the alert is coincident with a major volume change. The volume
168 at a particular date can be determined by hovering the mouse over the
169 appropriate location on the alert line.
171 * Don't file if the alert is on a graph with very low volume (< ~200 data
172 points); it's probably noise, and we probably don't care even if it isn't.
174 * Don't file if the graph is really noisy (but eyeball it to decide if there is
175 an underlying important shift under the noise).
177 * Don't file if the alert is in the "Known Ignorable" list:
178 * SimpleCache on Windows
179 * DiskCache on Android.
181 For each Gasper alert, respond to chrome-network-debugging@google.com with a
182 summary of the action you've taken and why, including issue link if an issue
185 ## Investigating crashers
187 * Only investigate crashers that are still occurring, as identified by above
188 section. If a search on go/crash indicates a crasher is no longer occurring,
191 * On Windows, you may want to look for weird dlls associated with the crashes.
192 This generally needs crashes from a fair number of different users to reach
194 * To get a list of loaded modules in related crash dumps, select
195 modules->3rd party in the left pane. It can be difficult to distinguish
196 between safe dlls and those likely to cause problems, but even if you're
197 not that familiar with windows, some may stick out. Anti-virus programs,
198 download managers, and more gray hat badware often have meaningful dll
199 names or dll paths (Generally product names or company names). If you
200 see one of these in a significant number of the crash dumps, it may well
202 * You can also try selecting the "has malware" option, though that's much
203 less reliable than looking manually.
205 * See if the same users are repeatedly running into the same issue. This can
206 be accomplished by search for (Or clicking on) the client ID associated with
207 a crash report, and seeing if there are multiple reports for the same crash.
208 If this is the case, it may be also be malware, or an issue with an unusual
209 system/chrome/network config.
211 * Dig through crash reports to figure out when the crash first appeared, and
212 dig through revision history in related files to try and locate a suspect CL.
213 TODO(mmenke): Add more detail here.
215 * Load crash dumps, try to figure out a cause. See
216 http://www.chromium.org/developers/crash-reports for more information
218 ## Dealing with old bugs
220 * For all network issues (Even those with owners, or a more specific labels):
222 * If the issue has had the Needs-Feedback label for over a month, verify it
223 is waiting on feedback from the user. If not, remove the label.
224 Otherwise, go ahead and mark the issue WontFix due to lack of response
225 and suggest the user file a new bug if the issue is still present. [Use
226 this issue tracker query for old Needs-Feedback
227 issues](https://code.google.com/p/chromium/issues/list?can=2&q=Cr%3AInternals-Network%20Needs=Feedback+modified-before%3Atoday-30&sort=-modified).
229 * If a bug is over 2 months old, and the underlying problem was never
230 reproduced or really understood:
231 * If it's over a year old, go ahead and mark the issue as Archived.
232 * Otherwise, ask reporters if the issue is still present, and attach
233 the Needs-Feedback label.
235 * Old unconfirmed or untriaged Cr-Internals-Network issues can be investigated
236 just like newer ones. Crashers should generally be given higher priority,
237 since we can verify if they still occur, and then newer issues, as they're
238 more likely to still be present, and more likely to have a still responsive