Closed
Bug 29370
Opened 25 years ago
Closed 24 years ago
Cache doesn't work for CSS and JS files
Categories
(Core :: Networking: Cache, defect, P3)
Core
Networking: Cache
Tracking
()
VERIFIED
FIXED
mozilla0.9
People
(Reporter: pierre, Assigned: gordon)
References
()
Details
(4 keywords, Whiteboard: [nsbeta2-][nsbeta3-] [rtm-] hit during nsbeta2 standards compliance testing)
Attachments
(1 file)
(deleted),
application/x-gtar
|
Details |
Tested with latest debug and optimized builds on Mac and Windows. This bug is
much more reproduceable on Mac than on Windows, and more reproduceable on
optimized builds than on debug builds. I recommend debugging it on the Mac.
- Go to http://www.w3.org/Style/CSS/
- Click reload again and again
==> You end up getting the page displayed without one or several stylesheets. Of
course, it's easier to spot when none of the sheets is loaded because the page
shows as plain HTML but keep an eye and you will notice sooner or later that you
can have some of the sheets loaded but not the others, especially when it
displays the message at the top "(This page uses CSS style sheets)" because the
class "hide" is missing.
Another way that allowed to reproduce the problem systematically with an
optimized build on the Mac is:
- Go to http://mvideo.cjb.net/
- Click on "2000" then on "Feb" then on one of the dates
==> You can see that some of the texts are displayed with the nifty font and
correctly centered while some other ones appear in Times and left-aligned.
Assigned to attinasi because it may be a side-effect of bug 17309 (although I
think I heard about sporadic problems with loading stylesheets a bit before the
fix for bug 17309 was in). CC'd Necko people who may have a clue, or at least
help Marc to set some breakpoints at the correct places.
Reporter | ||
Comment 1•25 years ago
|
||
Nominating for beta1. Failing to load the stylesheets is fairly bad.
Keywords: beta1
Updated•25 years ago
|
Status: NEW → ASSIGNED
Reporter | ||
Comment 3•25 years ago
|
||
Bug 24819 was PDT- because it was difficult to reproduce. This bugs shows that
the problem is more common than originally thought, especially on the Mac.
Comment 5•25 years ago
|
||
Looking at this one: first I have to duplicate it, then find out who is failing
and why, then determine a fix. SWAG at fix data: 3/1/00 (will update as I get
more info)
Whiteboard: [PDT+] → [PDT+] SWAG fix date: 3/1
Comment 6•25 years ago
|
||
I have had no luck reproducing this bug yet.
Pierre, when the problem shows up do you see anything on the console? There is a
printout in CSSLoaderImpl::DidLoadStyle that should dump to stderr when there is
an error loading the stylesheet:
if (mDocument && NS_FAILED(aStatus)) { // still have doc, must have failed
// Dump error message to console.
char *url;
aLoadData->mURL->GetSpec(&url);
cerr << "CSSLoaderImpl::DidLoadStyle: Load of URL '" << url
<< "' failed. Error code: " << NS_ERROR_GET_CODE(aStatus) << "\n";
nsCRT::free(url);
}
URLKey key(aLoadData->mURL);
Cleanup(key, aLoadData);
}
I am wondering if you see this, or if you can put an alert of some sort (or
breakpoint) there and try it on the mac (I have no mac to test on). I'll keep
trying to reproduce it.
Comment 7•25 years ago
|
||
I am unable to duplicate this on NT after several tedious hours of trying over
various link-speeds. Could this be a platform-specific networking issue? I'll
have to get some help on Mac or it will take me a lot longer as I have no Mac of
my own... Any offers?
Comment 8•25 years ago
|
||
I can reproduce this easily on Linux (cvs build from today). I'm on a 56K
line shared amongst 10 people (very slow).
Comment 9•25 years ago
|
||
Rick: Could this be a socket transport timeout issue? What's the timeout?
Does anyone know what the error code is that occurs?
Reporter | ||
Comment 10•25 years ago
|
||
I'll work on it. I can easily reproduce the problem and I heard that Mark is out
sick today.
Comment 11•25 years ago
|
||
Adjusting title of bug to indicate that this is a REload problem.
Adding notation that this will become PDT- if not resolved by 3/3. We'll have
to release note it.
Whiteboard: [PDT+] SWAG fix date: 3/1 → [PDT+] SWAG fix date: 3/1 w/b minus on 3/3
Updated•25 years ago
|
Summary: Stylesheets are not always loaded → Stylesheets are not always RE-loaded
Comment 12•25 years ago
|
||
I have a page on my local lan that does this quite often.
It won't ever do it the first time I visit it, but after that
first time, it happens most every time. I don't have to hit reload
for the bug to occur, I'm just revisiting the page,
so I'm thinking it doesn't have anything to do
with reloading a page in particular, but rather loading a page that has
been loaded once in the past. So perhaps it has something to do with the
page being fetched from the server and the css being loaded from the
memory cache. Also, the page contains many tables, so
it may be caused by it taking so long to layout all the tables.
Being that it's on a 10 mbs lan, I don't think it has anything to do
with network speen either.
Comment 13•25 years ago
|
||
Some test cases, in case they can be of help:
http://www.orakel.ntnu.no/~oyvindmo/csstest1.html
http://www.orakel.ntnu.no/~oyvindmo/csstest2.html
http://www.orakel.ntnu.no/~oyvindmo/csstest3.html
http://www.orakel.ntnu.no/~oyvindmo/csstest4.html
http://www.orakel.ntnu.no/~oyvindmo/csstest5.html
The test cases are:
1) <link ...> to static CSS file
2) Embedded CSS in <style>...</style>
3) <link ...> to PHP generating CSS identical to case 1
4) <link ...> to PHP generating CSS with Last-Modified in the future
5) <link ...> to PHP generating CSS with a 3 second delay in the middle
(The CSS is returned with Content-type text/css in all cases, by the way.)
Comment 14•25 years ago
|
||
how is the style sheet loaded? Pierre? do we load these URL's synchronously?
What api are we using?
Comment 15•25 years ago
|
||
My own observation, after finally getting to run my test cases myself, is that
the only case that gets the bug to manifest itself, is csstest1.html.
Looking at case 1 and 3, the difference is in the HTTP headers that get sent
with the style sheet. Here is the diff of how it looks when fetching the style
sheets with HTTP/1.0:
--- httplog-default.css Fri Mar 3 09:15:08 2000
+++ httplog-default.php3 Fri Mar 3 09:15:44 2000
Server: Apache/1.3.6 (Unix) PHP/3.0.8
-Last-Modified: Wed, 09 Feb 2000 14:06:46 GMT
+Last-Modified: Wed, 09 Feb 2000 14:06:47 GMT
Connection: close
Content-Type: text/css
-ETag: "36ae4c-52c-38a17476"
-Accept-Ranges: bytes
-Content-Length: 1324
-X-Pad: avoid browser bug
I've tried reloading csstest3.html hundreds of times, but the bug doesn't
appear there, it seems.
Reporter | ||
Comment 16•25 years ago
|
||
jar: this is not a RE-load problem, it is a page load problem.
valeski: the nsHTMLContentSink calls mCSSLoader::LoadStyleLink() which calls
CSSLoaderImpl::LoadSheet() which loads the style sheet synchronously by using
NS_OpenURI() and a nsIUnicharInputStream.
Summary: Stylesheets are not always RE-loaded → Stylesheets are not always loaded
Comment 17•25 years ago
|
||
Feels like a dupe of http://bugzilla.mozilla.org/show_bug.cgi?id=26297
Comment 18•25 years ago
|
||
After another hour of trying to dup this one I finally did get a single failure.
The console had the following error message printed out:
CSSLoaderImpl::DidLoadStyle: Load of URL 'http://www.w3.org/Style/map-ns.css'
failed. Error code: 2
I cannot get it to happen again.
Is anybody else seeing this error message when the problem occurs?
Reporter | ||
Comment 19•25 years ago
|
||
*** Bug 30308 has been marked as a duplicate of this bug. ***
Comment 20•25 years ago
|
||
Transitioning to PDT-. This bug seems really hard to reproduce... and we have
to get out the beta.
Whiteboard: [PDT+] SWAG fix date: 3/1 w/b minus on 3/3 → [PDT-] SWAG fix date: 3/1
Comment 21•25 years ago
|
||
(adding myself to cc list)
Just came to report this with our site actually and found it listed in the very
first message.
Some people seem to have a hard time duplicating this. However, this doesn't
seem very sporratic at all. It happens absolutely every time at
http://mvideo.cjb.net/ ...
The main frame doesn't load the CSS.
The frame with the navigational ".2000." doesn't load the CSS, nor do any of
the navigational popups with months/weeks under it. They should all be centered,
in different colors, and a different font.
This happens absolutely every time, i'm using 2000030508 on Windows 98SE ...
Comment 22•25 years ago
|
||
I have some test perl cgi scripts at
http://www.vortxweb.net/ccs-pg/my-cgi/toykar/chayka5.cgi
There are 4 scripts. Submitting from one page initiates the next. The external
stylesheets are loaded only on the first visit to the first page. Pages 2-4 get
no styles, and reloading does not correct it. It happens every single time. I'm
running Win98 build 2000030215.
Comment 23•25 years ago
|
||
I think I've found a link between the stylesheet loading problem and enabling
the memory cache. I have a lot of documents using the same stylesheet. When
memory cache is disabled (though prefs), stylesheets load fine. However,
whenever I enable memory cache, stylesheets stop loading intermittently,
specially for the longer pages. And once it stops loading, there is no way to
load it with the other pages (using the same style sheet) as well. Disabling
memory cache enables the style sheet again.
Comment 24•25 years ago
|
||
One word of warning - disabling memory cache and re-enabling it (both through
the prefs dialog) crashes mozilla with a stack overflow error in
nsTextFormatter::vsprintf_append (WinNT 4 SP 6a) - filing separate bug for this.
Comment 25•25 years ago
|
||
Confirming on 2000030508/win98se that disabling the Mem Cache allows style
sheets to work properly.
It appears that style sheets *do* work properly in a non-framed page whether or
not the cache is enabled. But in a framed page they only seem to load when the
mem cache is disabled.
(also note, I do not get the crash you spoke of disabling/enabling mem cache)
Comment 26•25 years ago
|
||
Reporter | ||
Comment 27•25 years ago
|
||
I confirm that the problem is related to the Mem Cache.
Changed Component to Networking:Cache and reassigned to owner.
Assignee: attinasi → gordon
Status: ASSIGNED → NEW
Component: Style System → Networking: Cache
QA Contact: chrisd → tever
Reporter | ||
Comment 28•25 years ago
|
||
Cleared the Status Whiteboard to ask the PDT team to reconsider.
Gordon: how bad would it be to ship with the Mem Cache turned off by default?
Whiteboard: [PDT-] SWAG fix date: 3/1
Comment 29•25 years ago
|
||
Jud asked me to look at this bug...
Based on the fact that the problem is with frames pages, I'm making a WAG that
more than one frame is using the same style sheet and is exposing a bug that
occurs only when a URL that is being loaded into the cache is requested a second
time *before* the first load completes.
At one time, I was going to have the cache code coalesce such overlapping
requests into a single network access, but after a Necko meeting in which we
discussed the level of software complexity imposed by this feature, the team
decided that, for now, the first request would access the cache and subsequent
requests for the same URL would bypass the cache until the first request
completed.
The code in nsHTTPChannel::CheckCache() is responsible for enforcing this
no-coalesce policy. In particular, the code that checks whether or not the
cache entry is being updated may be at fault:
http://lxr.mozilla.org/seamonkey/source/netwerk/protocol/http/src/nsHTTPChannel.
cpp#772
Another even more likely culprit, in my mind, is the code in
nsHTTPHandler::NewChannel() that attempts to coalesce requests for the same URL
into the same channel. I worry about that because there are a bunch of
nsHTTPChannel instance variables maintained by the HTTPChannel code and there
are likely to be assumptions that the state is per-connection, not per-URL. I
would try disabling this nsHTTPChannel-coalesce code and see if the bug goes
away.
Incidentally, one way to reproduce this problem without the complexity of frames
might be to create an HTML page which has many copies of the same *large* image
on it, but each with different WIDTH and/or HEIGHT attributes. (This will cause
many requests for the same image URL - I haven't tried this myself.)
Comment 30•25 years ago
|
||
I'd like to clarify that this bug doesn't happen on just pages with frames.
In the original test cases, none of the pages have frames. It seems to me
that it has something to do with loading a large page where the stylesheet
document is already in the memory cache. I can get the bug to occur on
almost any such page.
The page for which I see this almost everytime I load the page is where
I have an index page with many anchored links into a second page. After
the first access to the second page, each access from the index page to
the second page causes this bug to occur.
It has sensitive information on it, so I can't post it here, but I'll see
if I can get a good test case built from it.
Comment 31•25 years ago
|
||
OK, I've got a test case that will allow me to reproduce this bug every time.
The occurance of this bug seems to depend on when things are loaded, as I
am able to reproduce on this particular testcase over my lan, and off of
localhost, but not over the internet.
To reproduce:
1) set up a web server on your local machine
2) download the testcase i'll be attaching to my next comment
3) gunzip and untar it where you can access it from your webserver on localhost
4) access css-test/index.html with mozilla, IE http://localhost/css-test/
5) Click on one of the numbers to load the page with stylesheets
6) Click back
7) Click on a different number to load the page with stylesheets a second time
Result:
On step 7, the page will load without stylesheets
Tested on a Pentium II 400 w/ 64 MB ram.
Running Debian Woody and apache 1.3.9-11 (default debian config).
Comment 32•25 years ago
|
||
Comment 33•25 years ago
|
||
This URL may also produce the same results, depending on your line speed:
http://ss-i.com/~dan/test. This URL works fine for me, but I think it's
because of my slow connection.
I also at times get a partial stylesheet. Only parts of the stylesheet
elements are applied. I'll get a blue background only, or a blue
background and white boxes, for example.
This bug has also been reproduced on win32 2000030516.
Comment 34•25 years ago
|
||
*** Bug 29231 has been marked as a duplicate of this bug. ***
Comment 35•25 years ago
|
||
*** Bug 28805 has been marked as a duplicate of this bug. ***
Comment 36•25 years ago
|
||
*** Bug 29732 has been marked as a duplicate of this bug. ***
Comment 37•25 years ago
|
||
I think bug 28297 is a duplicate of this one. Although that bug is about
external JS files not loading and this one is about external CSS files not
loading, the descriptions are otherwise very similar. Both problems involve
external files sourced into the HTML page, both problems go away when the memory
cache is disabled, both problems are hard to reproduce for some people but easy
for others, and both problems have test cases that work consistently when
someone is running a web server on the same workstation on which they are
running Mozilla.
Comment 38•25 years ago
|
||
*** Bug 28297 has been marked as a duplicate of this bug. ***
Comment 39•25 years ago
|
||
*** Bug 28643 has been marked as a duplicate of this bug. ***
Comment 40•25 years ago
|
||
Putting on PDT- radar for beta1. No time to address for beta1.
Whiteboard: [PDT-]
Comment 41•25 years ago
|
||
Can we at least get external CSS style sheets and external JS files to NOT be
cached for beta1? Also, can anyone explain why caching CSS or JS file should be
any less reliable than caching HTML or image files? Should we turn off caching
altogether for beta1?
Comment 42•25 years ago
|
||
This bug severely affects all of the netcenter sidebar customization pages
(weather, news, etc). You get javascript errors up the yazoo, since it's looking
for functions that are defined in external js files. It also makes customizing
impossible since the files that don't load are necessary.
To see this simply:
1. Open sidebar
2. Click on customize
3. Click on Favorites and add a News panel
4. Click on customize
Results: you will get js errors and the list of news sources will not load. If
it works the first time, simply repeat
This problem goes away when you disable the memory cache. Yet again
renominating, and changing the summary to include js files.
cc:ing rginda
Summary: Stylesheets are not always loaded → Stylesheets and JS files are not always loaded
Whiteboard: [PDT-]
Comment 43•25 years ago
|
||
That bug seems to destroy layout of a lot of sites, all sites reusing external
css/js for me in recent builds.
I can't reproduce it with M14, though.
That all is happening for me on win98 & a fast internet connection...
Comment 44•25 years ago
|
||
*** Bug 30899 has been marked as a duplicate of this bug. ***
Comment 45•25 years ago
|
||
In an attempt to bypass the mem cache (for now) for css urls we'll try
bypassing the cache. Can someone who can reproduce this please try the following
change?
Index: nsCSSLoader.cpp
===================================================================
RCS file: /cvsroot/mozilla/layout/html/style/src/nsCSSLoader.cpp,v
retrieving revision 3.35
diff -u -r3.35 nsCSSLoader.cpp
--- nsCSSLoader.cpp 2000/02/02 23:04:27 3.35
+++ nsCSSLoader.cpp 2000/03/08 01:25:27
@@ -1116,7 +1116,8 @@
nsCOMPtr<nsILoadGroup> loadGroup;
mDocument->GetDocumentLoadGroup(getter_AddRefs(loadGroup));
- result = NS_NewStreamLoader(&loader, urlClone, aData, nsnull, loadGroup
);
+ result = NS_NewStreamLoader(&loader, urlClone, aData, nsnull, loadGroup
,
+ nsnull, nsIChannel::FORCE_RELOAD);
#ifdef NS_DEBUG
mSyncCallback = PR_FALSE;
#endif
Comment 46•25 years ago
|
||
Can someone have a look at bug 28490 ? I suspect that it might be a dup of
this. In both cases, the problem disappears after disabling memory cache, and
both are problems with CSS not being loaded / loaded partially.
Reporter | ||
Comment 47•25 years ago
|
||
*** Bug 28490 has been marked as a duplicate of this bug. ***
Reporter | ||
Comment 48•25 years ago
|
||
Unless someone beats me to it, I'll try the fix proposed by valeski tomorrow (my
build is broken right now).
Comment 49•25 years ago
|
||
*** Bug 26538 has been marked as a duplicate of this bug. ***
Comment 50•25 years ago
|
||
Hmm... in my builds as of 3/06/2000 I can no longer reproduce this (I still can
in M14). Looking through Bonsai, I didn't see any obvious reason why this would
be fixed...
warren made a huge checkin that touched caches, CSS, network code, the whole
nine yards that was supposedly implementing thread-safe addref/release
(b=21556). It was done to prevent crashes on SMP but I wonder... espescially
since this was an itermitant, timing-sensitive crash.
Either that or my debug builds can't reproduce it becasue of changed timing.
Comment 51•25 years ago
|
||
Ok, valenski's patch fixed M14. However, I can't reproduce anymore with or
without it on the tip (as of 03/06). So the bug may have been fixed in passing
by some other change... or it may just have been buried by some timing change.
Comment 52•25 years ago
|
||
This is dogfood gor QA automation. But we will try todays build.
Keywords: dogfood
Comment 53•25 years ago
|
||
Per desale -"Can not reproduce it with todays builds.
Looks like it got fixed magically." Marking Verified.
Status: NEW → RESOLVED
Closed: 25 years ago
Resolution: --- → FIXED
Comment 54•25 years ago
|
||
I'm trying to understand what the status is on this one. puetzk said he
couldn't reproduce this now, but is that with or without valeski's fix? Did the
fix/hack go into the tree (to not cache stylesheets)?
Comment 55•25 years ago
|
||
I just downloaded 2000030808 / Win98 and this problem is in absolutely no way
fixed.
Everything at mvideo.cjb.net should be loading info from the linked CSS file,
and nothing appears to be doing it.
Once again, disabling the cache makes them load fine ...
This is not fixed.
Comment 56•25 years ago
|
||
Reopening.
sentinel: Can you try valeski's hack? Thanks,
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Comment 57•25 years ago
|
||
^ = Stupid Non-Programmer Windows User. :)
I could if someone told me what to do.
Comment 58•25 years ago
|
||
please try my above patch.
Comment 59•25 years ago
|
||
Like I said, I would if I knew how ... :)
I'm not a programmer and I have no idea what to do with that snippet of code ...
Comment 60•25 years ago
|
||
I am no longer able to reproduce this problem on the tip (I don't have
valenski's patch on my tip build). However, I am able to reproduce quite
consistently on my M14 build. Applying valekski's hack to my M14_BRANCH build
makes it impossible to reproduce this bug with it (ie, succesfull avoids the
problem). I think some people are still able to reproduce this with tip builds,
but I can't anymore. I have no idea why.
Comment 61•25 years ago
|
||
I can still reproduce this bug on my cvs build, but not on a downloaded
nightly (2000030709). And now, after applying valeski's patch, I am no
longer able to reproduce on my cvs build. I think it did the trick.
But I think this only fixes half the problem.
Are people still able to reproduce the javascript half of the problem?
Summary: Stylesheets and JS files are not always loaded → External stylesheets and JS files are not always loaded
Comment 62•25 years ago
|
||
*** Bug 31060 has been marked as a duplicate of this bug. ***
Comment 63•25 years ago
|
||
This happens consistently for me on Linux nightly build 2000.03.08.09: a style
sheet is only used when first loaded, not on reload, not on other pages
referencing the same stylesheet. Bug 31031 is probably a dup of this one, it's
about JS, and was filed today, so it's not fixed for JS either.
Summary: External stylesheets and JS files are not always loaded → Stylesheets and JS files are not always loaded
Comment 64•25 years ago
|
||
The same thing seems to happen with XUL. Here are comments from our 3rd party
partner:
"if you reload an xul file that calls a stylesheet, the browser refreshes only
the xul file but not the css
file containing the stylesheet. The problem occurs if the css file is called
from a web server, but not if it's loaded locally"
Comment 65•25 years ago
|
||
zach: Can you try the build with the patch?
To apply it, all you have to do put the patch in a file (e.g. foo) and do this:
cd layout/html/style/src
patch -c <foo
make
Comment 66•25 years ago
|
||
Oops zack: Take out the -c in the patch line, above.
Comment 67•25 years ago
|
||
Sorry, I'm strictly a nightly build kind a guy.
Comment 68•25 years ago
|
||
Putting on PDT+ radar for beta1. Must fix by 03/10 to make beta train.
Whiteboard: [PDT+] w/b minux on 03/10
Reporter | ||
Comment 69•25 years ago
|
||
*** Bug 26736 has been marked as a duplicate of this bug. ***
Updated•25 years ago
|
Target Milestone: M14
Reporter | ||
Comment 70•25 years ago
|
||
The bug could still easily be reproduced with the tip on the Mac. Valeski's patch
fixed it so I got the approval and checked it in. I'm not closing the bug though,
because the absence of cache for Stylesheets and JS files may hurt the
performance (see bug 17309).
Changed the title to "Memory Cache doesn't work for CSS and JS files".
Set target to M15. Reduced the CC list. Cleared the Status Whiteboard. Added
'perf' keyword.
Comment 72•25 years ago
|
||
I'm not familiar with the code, but appears to me jud's fix only addresses css
files, can someone confirm?
Reporter | ||
Comment 73•25 years ago
|
||
See bug 28297 for a description of the problems with JS files. I can't reproduce
the bug described there with my.netscape.com.
paulmac: could you try too?
Comment 74•25 years ago
|
||
If people are still having problems reproducing this error, it happens readily
at http://www.w3cdom.com/front.html
After the page loads, click the "WELCOME" menuitem. This attempts to reload the
iframe to its inintial state. Instead, it reloads without the stylesheet. If the
main page is then RELOADED, it loses the css and javascript files everytime.
Disabling mem-cache fixes this.
I'll try to make a simple test-case from this...
Reporter | ||
Comment 75•25 years ago
|
||
*** Bug 31202 has been marked as a duplicate of this bug. ***
Assignee | ||
Comment 76•25 years ago
|
||
Assigning to Jud, since he has an idea for how to handle JS files.
Assignee: gordon → valeski
Status: REOPENED → NEW
Comment 77•25 years ago
|
||
I can still reproduce the problems from bug 28297 on 030914 builds by reloading
http://my.netscape.com/setup_frameset.tmpl?services=weather&mn_yes=1
and
http://www.zapogee.com/~myk/tc30.html
These are both problems with js files reloading which go away with mem cache
turned off. However, for whatever reason, I can't reproduce it on Jud's debug
build. I think he should check it in, however, since the css patch seems to have
worked. Remarking PDT+, because it was removed in error.
Whiteboard: [PDT+]
Comment 78•25 years ago
|
||
I'm going to do an optimized build and kick it off to paulmac to ensure that my
fix works for js files (since I can't reproduce the problem on my machine)
updated status to reflect expected check-in date
If anyone is still having problems with css files on builds 3/9 or after please
speak up.
Whiteboard: [PDT+] → [PDT+] will check in 3/10 if fix works on optimized
Comment 79•25 years ago
|
||
running build 2000030913 in WinNT4.0...
External CSS problem seems resolved, javascript problem is till there...
To see:
1- make sure mem-cache is on...
2- go to http://w3cdom.com/lab/bugwatch.shtml
3- RELOAD page...
External javascript fails and will not come back till restart of browser or
mem-cache disable.
Comment 80•25 years ago
|
||
valeski just checked in workaround for js files, someone please verify with 3/10
builds, thanks
removing pdt+ but leaving bug open and re-assigning to gordon
Assignee: valeski → gordon
Whiteboard: [PDT+] will check in 3/10 if fix works on optimized → workaround checked in
Comment 81•25 years ago
|
||
*** Bug 31046 has been marked as a duplicate of this bug. ***
Comment 82•25 years ago
|
||
PDT would like this bug closed out and a new one opened please.
Summary: Memory Cache doesn't work for CSS and JS files → [PDT-]Memory Cache doesn't work for CSS and JS files
Comment 83•25 years ago
|
||
thanks jud, this worksforme now with the js testcases listed above. I'm going to
leave this bug open, but remove the dogfood keyword so the PDT team won't bug us
anymore :-)
Keywords: dogfood
Summary: [PDT-]Memory Cache doesn't work for CSS and JS files → Memory Cache doesn't work for CSS and JS files
Comment 84•25 years ago
|
||
2000031008 (NT4 128mb)
External css and js now seem to work with cache enabled. (on my testcase at
least..)
You folks rock!
Reporter | ||
Comment 85•25 years ago
|
||
I confirm: the JS testcases above work in today's build.
Marking Fixed.
Status: NEW → RESOLVED
Closed: 25 years ago → 25 years ago
Resolution: --- → FIXED
Comment 86•25 years ago
|
||
Only a workaround has been checked in, so this bug shouldn't be marked fixed, in
my opinion. Re-opening bug since the original problem still remains and will
hopefully get fixed post-beta.
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Reporter | ||
Comment 87•25 years ago
|
||
*** Bug 29929 has been marked as a duplicate of this bug. ***
Reporter | ||
Comment 88•25 years ago
|
||
*** Bug 32704 has been marked as a duplicate of this bug. ***
Comment 89•25 years ago
|
||
I think bug 32704 is just bad browser detection, nothing to do with caching,
that I can see.. just plenty of .layers/.all code.
bug 29929 - Cannot reproduce with or without cache set on or off. I increased
amout of text in <dl> to 100 times what is at URL, and it will still not drop
the stylesheet.
I don't think either of these is an actual bug, much less a dupe of this one...
(all testing done with build 2000032115, WinNT4, 128mb)
FWIW, external .js and .css have been behaving fine (for me) since 3/10's "fix"
(tested with nearly everyday's win-build since 3-10-2000)
Reporter | ||
Comment 90•25 years ago
|
||
*** Bug 33036 has been marked as a duplicate of this bug. ***
Comment 91•25 years ago
|
||
When the actual fix is in we need to remember to go back and re-enable the
caching in the CSSLoader (and in the JS counterpart too I assume).
Comment 92•25 years ago
|
||
*** Bug 17309 has been marked as a duplicate of this bug. ***
Comment 93•25 years ago
|
||
Moving to M17 which is now considered part of beta2.
Target Milestone: M16 → M17
Keywords: beta2
Whiteboard: workaround checked in → workaround checked in, ? days
Comment 94•25 years ago
|
||
REassign. Putting in a big number since I think this is hard
Assignee: gordon → davidm
Status: REOPENED → NEW
Keywords: beta2
Whiteboard: workaround checked in, ? days → workaround checked in, 5 days
Comment 95•25 years ago
|
||
Putting on [nsbeta2-] radar. Adding nsbeta3 keyword for performance work in PR3.
Keywords: nsbeta3
Whiteboard: workaround checked in, 5 days → [nsbeta2-]workaround checked in, 5 days
Comment 96•25 years ago
|
||
David has left netscape and davidm@netscape.com is an invalid email address.
Assigning his five remaining bugs to gagan to find a new owner for them. Thanks
Comment 97•25 years ago
|
||
oops i didn't actually reassign to gagan last time. sorry for the spam
Assignee: davidm → gagan
Comment 98•25 years ago
|
||
*** Bug 44461 has been marked as a duplicate of this bug. ***
Comment 100•25 years ago
|
||
This bug is still existing. I can reproduce it very easy with my local
webserver.
Comment 101•25 years ago
|
||
neeti could you investigate this? Considering its effect in performance I am
making this a nsbeta3+
Whiteboard: [nsbeta2-]workaround checked in, 5 days → [nsbeta2-][nsbeta3+]workaround checked in, 5 days
Comment 102•25 years ago
|
||
See bug 46548 -- css and js files loaded multiple times. Possibly related.
Keywords: footprint
Comment 103•25 years ago
|
||
From re-reading the comments I have the impression that this bug-report is a
bundle of three related problems.
FIRST, there is the CSS cache problem mentioned in the summary. This lead me and
some authors of earlier comments to this bug. From my understanding, caching was
disabled as a workaround for problem #2 (see below). Not caching stylesheets can
lead to a very bad performance on sites working with large, global stylesheets
that are included on every page.
SECOND, there is a problem with stylesheet loading. It seems to be timing
related and manifests itsself in broken page styles (the stylesheet is either
not loaded completly or not loaded at all). This problem can be reproduced best
when using a local web-server, but other comments indicate that it also happens
on remote web-servers if there is a high bandwidth connection to the server. My
own experience (local site with frames) is, that this problem (let's call it
half-loaded style sheets) has a probability of 60%-80% - when the whole frameset
is reloaded. I have not seen it with unframed pages but previous comments
indicate that it also happens then, maybe just not that often. Half-loaded style
sheets occur completly independent of how the memory cache is configured. I can
reproduce it with all possible combinations of memory cache settings (ie 0kb &
disabled / xKb & disabled / yKb & enabled).
THIRD, there may or may not be a problem with javascript files. I have not
checked this, but there are other bug reports (XUL related like #46129) which
indicate, that this problem _also_ still exists.
My conclusions are:
--> Due to the independence of caching CSS and problem #2 (half loaded style
sheets), someone should probably reenable the caching of external style sheets.
--> Maybe it would be better to change the summary or break this bug up into
separate ones, because problem #2 (half-loaded style sheets) is not even
mentioned in the summary.
--> The whiteboard should be updated. In the moment, it reads "workaround
checked in, 5 days". What does it mean, when will this five day period end ?
Comment 104•25 years ago
|
||
changing whiteboard to remove the outdated info. the workaround was put in a few
months ago and it was simply to force reloads on css and js files.
Whiteboard: [nsbeta2-][nsbeta3+]workaround checked in, 5 days → [nsbeta2-][nsbeta3+]
Comment 105•25 years ago
|
||
When this is fixed, I will need to reexamine our behaviour on all the nasty tests
in the CSS ImportTest, especially the "infinite" ones:
http://www.bath.ac.uk/%7Epy8ieh/internet/importtest/
Keywords: testcase
Whiteboard: [nsbeta2-][nsbeta3+] → [nsbeta2-][nsbeta3+] hit during nsbeta2 standards compliance testing
Comment 106•24 years ago
|
||
BTW, "Pragama: no-cache" and "Cache-Control: must-revalidate" seem to be always
sent when fetching image.
Comment 107•24 years ago
|
||
nsbeta3-
We are not going to be able to get to this. Maybe we should take this up for RTM
if we think the performance needs to be fixed that badly.
RTM marked so this doesn't get lost in the wilderness.
Keywords: rtm
Whiteboard: [nsbeta2-][nsbeta3+] hit during nsbeta2 standards compliance testing → [nsbeta2-][nsbeta3-] hit during nsbeta2 standards compliance testing
Comment 108•24 years ago
|
||
not going to happen for rtm
Whiteboard: [nsbeta2-][nsbeta3-] hit during nsbeta2 standards compliance testing → [nsbeta2-][nsbeta3-] [rtm-] hit during nsbeta2 standards compliance testing
Comment 109•24 years ago
|
||
This is a pretty serious bug to slip. I guess the message to developers is
"Netscape supports CSS, but don't really use it in a meaningful way."
Comment 111•24 years ago
|
||
*** Bug 60746 has been marked as a duplicate of this bug. ***
Comment 112•24 years ago
|
||
*** Bug 66859 has been marked as a duplicate of this bug. ***
Comment 113•24 years ago
|
||
#1 Gordon told me that memory cache is used for IMAP only, so then the summary
is at least confusing. So was the first step to take it out the Memory cache to
avoid further problems?
#2 it seems that someone made a hack around, forced reload, but that brakes rfc
2116 compliance, but that doesn't seems to bother you, even after almost a year now!
#3 this bug increases network traffic, so this hack around introduced an serious
performance issue.
#4 can you at least make a user_preference for the time being? This way we don't
have to reload all that files over and over again!
#5 do you have any idea what is causing the real problem behind this bug?
If I whas editors then I would like to know why Mozilla and Netscape are taking
so long to fix this problem! Almost a year now! Is that because Netscape = AOL?
And that AOL get paid for increasing internet traffic?
That would be a hell of a story, don't you think?
Comment 114•24 years ago
|
||
Adding Gerv to CC. Gerv, I think when the new newsgroups have been created, we
should add a mozilla.conspiracy-theory group.
Assignee | ||
Comment 115•24 years ago
|
||
The men in black suits have been dispatched. The situation will be under control
short...oops! Is this mic on?
Comment 116•24 years ago
|
||
Nominating it for nsbeta1 might produce results :-) JS and CSS files should be
cacheable.
Gerv
Keywords: nsbeta1
Comment 117•24 years ago
|
||
*** Bug 66858 has been marked as a duplicate of this bug. ***
Comment 118•24 years ago
|
||
So, now it's confirmed, memory cache = disk cache, right? Why else did you dup
that last bug?
Comment 119•24 years ago
|
||
Revising summary as this problem is not particular to the memory cache.
Summary: Memory Cache doesn't work for CSS and JS files → Cache doesn't work for CSS and JS files
Comment 120•24 years ago
|
||
That makes it crystal clear
Comment 121•24 years ago
|
||
Sorry for the spam, just adding to CC list.
Comment 122•24 years ago
|
||
targetting for new cache landing.
Assignee: neeti → gordon
Status: ASSIGNED → NEW
Target Milestone: M17 → mozilla0.9
Comment 123•24 years ago
|
||
*** Bug 74450 has been marked as a duplicate of this bug. ***
Comment 124•24 years ago
|
||
I'm seeing the opposite problem, with the 4/4 Linux build. I have an html page
that does <script src="file.js"></script>; I make a change to file.js and try to
reload the html page, but the changes don't show up on reload or shift-reload.
(I have my cache pref set to "compare every time", if that matters.)
Clicking "Clear memory cache" and "clear disk cache", then restarting, doesn't
help (so apparently it didn't actually clear anything).
However, moving aside the directories Cache and NewCache in my profile directory
did finally show me the new page.
Comment 125•24 years ago
|
||
clear memory/disk cache buttons are not hooked up to the new cache yet.
CSS/JS caching have been re-enabled - see bug 73905 - and the problem of
reload/shift-reload not working is a known problem, but i don't know if
a bug has been opened for this yet... would you mind opening a bug for this?
thanks!
Assignee | ||
Comment 126•24 years ago
|
||
The Clear cache buttons in the cache preference panel now work, and Darin tells
me the reload/shit-reload problems were fixed with his recent landing.
If there are new issues with CSS and JS reloads, we'll open new bugs.
Marking FIXED.
Status: NEW → RESOLVED
Closed: 25 years ago → 24 years ago
Resolution: --- → FIXED
Comment 127•24 years ago
|
||
Bug still shows on Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; 0.8.1)
Gecko/20010323.
CSS and JS files are not cached at all.
Do I need to reconfigure something?
This is an example HTTP request:
GET /v/css/standardMoz06.css HTTP/1.1
Cookie: SessionV=5ha66c3; StylesheetV=defaultStd,standardMoz06
Host: localhost
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; 0.8.1) Gecko/20010323
Pragma: no-cache
Cache-Control: no-cache
Accept: */*
Accept-Language: en
Accept-Encoding: gzip,deflate,compress,identity
Keep-Alive: 300
Connection: keep-alive
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Comment 128•24 years ago
|
||
Ouch... Sorry! Didn't check release date vs. post date.
Downloaded rv:0.9+ Gecko/20010506 and CONFIRMED bug is resolved for my application.
Cache works as expected (If-Modified-Since, etc.). Load speed of pages increases
dramatically on a (simulated) 56k line.
Please mark as FIXED again.
Sorry again, Mark
Comment 129•24 years ago
|
||
It's not my bug but I'm able to mark it fixed, so I do change it back to fixed :)
Status: REOPENED → RESOLVED
Closed: 24 years ago → 24 years ago
Resolution: --- → FIXED
You need to log in
before you can comment on or make changes to this bug.
Description
•