User:Eighty5cacao/misc/Computer maintenance to address elsewhere/archive


 * This is not a formal archive; do expect occasional (copy)editing.

Firefox

 * Search engine autodiscovery wrongly notifies of an unencrypted version of the same search engine that the user just used in https (e.g., on qrobe.it, this is probably because Firefox notices only the first of two  tags)
 * Other than that case, if the page was reached via a query from the search bar, but autodiscovery finds a search plugin nevertheless, consider warning the user that the existing search plugin is outdated...? This is tricky because we have to download the new search plugin to make sure it differs from the one already installed.
 * If you start typing in the URL bar and very quickly click on one of the results, before the Places service has caught up to your typing (meaning that not all of the typed characters have yet been taken into consideration in matching history entries), it gets interpreted like pressing the Enter key. This causes an error if the typed text is not a valid URL or domain name. (possibly fixed in Firefox 57?)
 * Type some stuff in the URL bar that will result in at least one history match; click and hold on one of the results, and press Esc while the mouse button is still down. This is treated like pressing Enter and can result in a malformed-URL error (TODO: I have all "search from the URL bar"-type features disabled; does a Firefox installation/profile on default configuration attempt a DNS lookup as usual, or does it only try to interpret the typed text as a full URL?) Needs retesting: My choice to disable autofill and my userChrome.css hacks to remove the "Visit" suggestion from the URL bar were likely contributing factors.

Adblock Plus
(possibly obsolete? was written pre-WebExtensions)


 * The subscription dependency notification system ( and   parameters) should have a way to recognize that Fanboy Complete List includes EasyList and is therefore an acceptable alternative

Privacy Badger

 * Ref: Official EFF download and documentation page


 * User:Eighty5cacao/misc/Privacy Badger
 * [Is this still an issue?] I think the clause "... and sometimes (though not always) users may understand that the price of an excellent free tool like Google's search engine is measured in privacy, not money" needs to be revised to be less pushy to users. Yes, I know Google is breaking their "Don't be evil" promise. The problem is, interpreting that clause to refer to all first-party tracking makes it sound like, "Until Privacy Badger implements fingerprinting countermeasures, the best way to prevent first-party tracking is to vote with your feet by avoiding the site in question entirely." Obviously, this isn't practical because privacy risks may overlap with legitimate work needs.
 * Entropy estimator and other functional changes:
 * Cap/downscale entropy of cookies with names containing words like,   etc. (and the "(wikiname)Session" cookies used by the MediaWiki software)
 * Do something about sites that are used frequently by the end user as first party (not including known offenders such as Facebook) - possibly yellowlist?
 * Should the digits 2-9 be given a higher estimate than 0 and 1? (NB: will be resolved in GitHub ticket #771)
 * Detect first-party, HTTPS-secured POST requests to user-visible forms (as opposed to requests made by JavaScript, etc.) as a heuristic to determine whether the user is a registered user of a site
 * Give lower weight to existing cookies (that are not being modified by an inequivalent Set-Cookie header) than new cookies
 * There ought to be a way to bypass the reload that would ordinarily be forced when the user changes policies - this is useful on pages with login sessions or other uses of POSTDATA, when the user knows the reload would fail for that reason, and s/he is willing to manually log out of the site and log back in to continue testing. (The current workaround is to close the tab while Privacy Badger's settings panel is still open; this is not a real solution, as Firefox still appears to begin reloading the page [i.e. send network requests? verify] before the tab closure is fully processed. This has not been tested in cases where POSTDATA is present; presumably the usual warning message would pop up.)
 * First noticed 20:13, 14 January 2015 (UTC), Firefox 35 stable, Privacy Badger 0.2.5, still happens inconsistently: Closing the tab while Privacy Badger's settings panel is still open after changing settings instead triggers a reload of whatever tab got selected after the close.

Subscriptions for Adblock Plus and uBlock

 * Fanboy's Anti-Facebook Filters should probably have dependency changed to Fanboy Social List (which already has EasyList as its dependency in the first place)
 * Something in Fanboy's Social Blocking List hides everything except the video itself on YouTube watch pages (tested in uBlock Origin shortly prior to this edit) (needs retesting; possibly intentional?)

uBlock Origin

 * (Known issue?) Malware Domains list needs to use the URL officially sanctioned by ABP, in order to support HTTPS
 * Dashboard → 3rd-party filters tab: Enabling a "built-in" filter subscription does not fetch an updated version of it when "Apply changes" is pressed (or it gets flagged as outdated for other reasons?)
 * User ("my") filters that are being ignored due to redundancy need to be flagged in the UI somehow; the count displayed on the 3rd-party filters tab isn't really sufficient for debugging
 * The element picker needs "wider"/"narrower" shortcut keys like those of Element Hiding Helper for Adblock Plus.
 * Filters with a  directive cause HTTPS Everywhere to report a redirect loop, in cases where HTTPS-E also wants to rewrite the URL (because it was http to begin with). The actual redirection within uBO still seems to happen fine. (Possibly obsolete after the move to WebExtensions, though I have not verified this because the WebExtensions UI for HTTPS Everywhere does not show any notification of redirect loops)
 * CNET.com currently has a CSP, which breaks some injection filters in the default filters.txt. Possibly an intentional anti-adblock measure? (TODO: Exactly what are the symptoms? My first guess proved to be a different issue.)

Websites
These are really tech-evangelism items that require contacting the webmaster. Some issues could be cosmetically patched using user scripts or stylesheets via Greasemonkey or Stylish, but that's beside the point.


 * PubMed: Why doesn't turning off activity recording persist past a browser session (or perhaps a day)? Perhaps the cookie is tied to an IP address, and other students turning activity recording back on turns it on for me too? Does the site interpret DNT in any way?
 * The TASVideos forum supports syntax highlighting for AviSynth scripts. It automatically hyperlinks command names to the AviSynth wiki, but unfortunately such links are currently broken (404) due to site changes. This is fixable though:  →   (Example post here). Is it worth complaining to AviSynth about their uncool URIs or only to TASVideos about the fix on their end?
 * Imgur (possibly obsolete?): Viewing a page for a "user submitted" image switches the preferred homepage display setting accordingly; is this intentional, and is there any recommended workaround that doesn't involve registering an account?
 * Warn the user whenever a viral→usersub pref change occurs, like the current warning on the homepage, except also on individual image pages; make sure that canceling the pref change does not prevent the user from viewing the image (it MAY redirect to the image file itself on  if that is the easiest to implement; an example of undesirable behavior would be redirecting to the imgur.com homepage)

The Cutting Room Floor (tcrf:)

 * The  interwiki prefix appears to be configured for absolute http.

Miraheze wikis

 * The anchor "Orain wikis" points here for historical reasons.

All The Tropes

 * Darth Wiki (which is a category on AtT) should not use italic titles by default, since not all pages in the category are works. The original reason for this was presumably the "Unpublished Works" subdivision, but AtT has since deleted those pages. (TODO: Check why this occurs; that is, from where is the CSS rule loaded or the HTML tags applied? Also check whether Sugar Wiki suffers from the same problem.)
 * The  interwiki prefix appears to be configured for absolute http. (Needs testing elsewhere on Miraheze)

YouTube

 * Excessive CPU usage caused by another tab or program can sometimes trigger the "Experiencing interruptions?" message even though the video has already finished loading. Is that message intended only to signal network performance problems, or is it intentional that it also catches [CG]PU issues?

Android

 * Are there apps for non-rooted devices that allow the following:
 * Automatically activate the system-wide "power-saving mode" setting if battery ≤ X%
 * Keep Wi-Fi on during sleep iff battery ≥ X%
 * The below would probably require root to work with a stock ROM, if they existed
 * Remove the following nag messages from Google Play Store:
 * Exceptions to the default policy for automatic updating of apps
 * Setting up a Google Wallet(?) payment method when none is associated with the Google Account
 * Make all developer options persist across reboots, or at least 4x MSAA
 * Do not wake screen when charge reaches 100%
 * Activate GPS iff no network connection is available

CCleaner for Windows
Why doesn't the "Hide warning messages" setting hide the warning messages that result when making changes to the Custom Files/Folders?

G'MIC

 * For the 64-bit Windows build, the supplied version of  causes some built-in GIMP plugins to fail, specifically ,  , and  . (I don't really need these in my workflow, but this is of course a reportable issue nonetheless.) Apparently they normally look for   somewhere else in the GIMP installation, but they give the working directory higher precedence in the path. (Resolved PEBKAC: I should have extracted G'MIC elsewhere and reconfigured GIMP to look in that folder. But it's probably still worth asking for the documentation to be clarified.)

Java plugin for Windows

 * Please do not open the browser-check page after upgrade if the default browser is a 32-bit application and only a 64-bit version of Java is installed.

MediaWiki software

 * AbuseFilter extension: Attempting to view a diff or oldid (properly " ") of an abuse filter that has only one revision wrongly displays a message, "You are editing an old version of this filter. The statistics quoted are for the most recent version of the filter. If you save your changes, you will overwrite all changes since the revision you are editing." This is wrong because the (single) revision is the newest; a better message would say something about the absence of a previous revision with which to compare. An example on this wiki is Special:AbuseFilter/history/15/diff/prev/41 (which actually redirects to Special:AbuseFilter/history/15/item/41) as of this edit. I have also noticed this on the English Wikipedia, though I have no corresponding example handy.
 * Special:LinkSearch should explicitly show a warning message that a path has been ignored when the domain contains a wildcard. For example, I once tried entering the query, but this is treated as simply  . (I know better than to do that now, but newbies who didn't RTFM might not.)
 * Using a pipe trick in the title of a new talk section (via the plus sign/ ) causes the edit summary and post-save redirect to use an incorrect section anchor.
 * If multiple sections on a page use the same title, saving an edit to one of those sections always redirects to the first section with that title. Adding the appropriate number e.g.,  , etc. to the   markup in the edit summary does not affect the post-save redirect, which scrapes the section title from the saved wikitext.
 * Moving wikipedia:Template:Anchor to a different position within a section heading causes the section not to be found when the edit is saved (the post-save redirect uses no anchor).

Notepad++
TODO: This behavior is not exclusive to HTTPS Everywhere rulesets nor to TIS-620, but I didn't take notes the last time I reproduced this issue. (The case I recently saw involves a character which StepMania displays as a star or full-width (Japanese) asterisk&mdash;don't remember which. The incorrect character set was something Western with an ISO/IEC number [?], but I don't have on hand the exact file or answers to these questions.)
 * ...misdetects some of my HTTPS Everywhere user ruleset files as having TIS-620 character encoding, even though they are really UTF-8 per the XML standard. (Hypothesis: something to do with the superscripted numerals ¹²³ and the right arrow character →, when the rest of the ruleset isn't long enough to support a claim of UTF-8) As expected, this behavior is present only when a BOM is absent.
 * Pressing Tab when a multi-line selection is present indents the selected lines. If there is not a preference or a keyboard shortcut to behave like a "dumb" text editor by simply replacing the selection with a tab, one should be added.
 * Please provide an option to disable the autofilling of the search query with the word surrounding or adjacent to the cursor.

Shark007 Advanced Codecs

 * Why does Firefox always wrongly complain about not being the default browser after uninstalling Shark007's Advanced Codecs? While Firefox's warning message is on screen, IE claims not to be the default browser either (checked only via Internet Settings, not by actually starting IE), and IE but not Firefox is listed in the Default Programs control panel. No other browsers are installed on my system. (This may be less of an issue now that current versions support upgrade installation)

Legacy
I assume Skype is reading back the audio output and looking for clipping, though as stated, there isn't enough clipping for me to hear. Is a little clipping really that bad? Does this relate to audio enhancements applied by the sound driver? Is there some misuse of the program-specific volume APIs in Windows Vista/7? ([//community.skype.com/t5/Windows-archive/Skype-6-always-showing-yellow-call-quality-bars-after-the-update/td-p/1307566 Someone on the official forums mentioned this])
 * Why does the call quality rating system penalize one out of five points for "speaker too loud," even though I don't hear any problem? It's been like this on every computer on which I've ever used Skype.
 * When the installer attempts to set Bing as the search engine in Firefox, it modifies  in a way that corrupts it and therefore triggers the "Invalidprefs.js" backup code from bmo:361102. Specifically, it adds spurious line breaks into pref lines that are longer than a certain length. Also, it attempts to change the pref   even though Firefox ≥36 uses a locale-specific pref name such as.
 * The Skype behavior needs to be re-verified (08:34, 10 March 2018 (UTC)), but the Firefox behavior may change soon in a way that reduces data loss; see bmo:107264 and developer Nicholas Nethercote's blog post about the new prefs parser.

Version 8

 * Moved to User:Eighty5cacao/misc/Computer maintenance to address elsewhere § Skype for Windows (version 8)

__NOINDEX__