On 07/17/2017 04:20 AM, Amos Jeffries wrote:
> On 17/07/17 06:14, Alex Rousskov wrote:
>> Since we will switch to PR-only commit
>> path (for master) as soon as Eduard's CI integration is ready, I suggest
>> that we keep the current review-required setup to practice. We can
On 07/16/2017 06:40 AM, Amos Jeffries wrote:
>> On 16/07/17 16:43, Alex Rousskov wrote:
>>> I tried to "lock" the
>>> official repository on Github to prevent commits (without pull
>>> requests), but I have not tested whether that lock was
On 07/15/2017 10:50 PM, Amos Jeffries wrote:
> The -5 and -4 repo's are going to be deleted ASAP, right?
Those branches are handled under issue #18:
https://github.com/squid-cache/squid/issues/18
I was going to hide them first, but I do not see to find a way to do
that -- I was probably thinking
On 07/11/2017 10:20 PM, Alex Rousskov wrote:
> 2017-07-11: No more new tags in the official bzr repo.
> 2017-07-13: No more new commits(*) in the official bzr repo.
> 2017-07-14: Migration starts.
> 2017-07-15: Anticipated optimistic migration end.
> 2017-07-18: Anticipa
On 07/14/2017 09:18 AM, Christos Tsantilas wrote:
> SslBump was ignoring origin server certificate changes and using the
> previously cached fake certificate (mimicking now-stale properties).
I suggest replacing the above claim with a more accurate one (based on
our off-list discussions):
"""
On 07/08/2017 04:30 PM, Alex Rousskov wrote:
> On 07/01/2017 10:16 PM, Alex Rousskov wrote:
>> On 07/01/2017 12:57 AM, Amos Jeffries wrote:
>>> On 01/07/17 10:44, Alex Rousskov wrote:
>>>>
>>>> If caching IPv6 /etc/hosts entries when IPv6 suppor
Hello,
The Squid Project will be transitioning its official source code
repository from bzr to git. Most of the currently stored official
information will be preserved during the migration. The lists of all bzr
branches and tags to be migrated or abandoned are available[1,2]. If you
find any
On 07/09/2017 04:30 AM, Amos Jeffries wrote:
> On 09/07/17 15:28, Alex Rousskov wrote:
>> On 06/10/2017 06:27 AM, Amos Jeffries wrote:
>>> - I have not gone through and renamed uses of SessionPointer beyond
>>> those directly involved with the above code shuffle.
o much controversy. The details are further below.
> On 27/04/17 05:24, Alex Rousskov wrote:
>> Needless to say, I would be happy if we can come up with better
>> definitions or even better concepts. The above is a starting point.
> I do not think it is up to us to define these
On 07/01/2017 10:16 PM, Alex Rousskov wrote:
> On 07/01/2017 12:57 AM, Amos Jeffries wrote:
>> On 01/07/17 10:44, Alex Rousskov wrote:
>>>
>>> If caching IPv6 /etc/hosts entries when IPv6 support is disabled was an
>>> accident, then I will remove that
On 07/08/2017 10:36 AM, Amos Jeffries wrote:
> The input buffer is no longer truncated when overly long. All callers
> have been checked that they handle the bool false return value in ways
> that do not rely on that truncation.
That is a useful change, thank you. I cannot check whether any
On 07/01/2017 12:57 AM, Amos Jeffries wrote:
> On 01/07/17 10:44, Alex Rousskov wrote:
>>
>> If caching IPv6 /etc/hosts entries when IPv6 support is disabled was an
>> accident, then I will remove that functionality, simplifying the patch.
>> If it was intentional,
On 06/30/2017 10:07 PM, Alexandr wrote:
> В Пт, 30/06/2017 в 11:52 -0600, Alex Rousskov пишет:
>> Please keep our terms confidential.
So much for that. Sigh. You should be more careful.
Alex.
___
squid-dev mailing list
squid-dev@lists.squid-
On 06/30/2017 10:07 PM, Alexandr wrote:
> В Пт, 30/06/2017 в 11:52 -0600, Alex Rousskov пишет:
>> Measurement Factory can add basic upstream proxy support for SslBump in
>> Squid v5. That support will match existing upstream proxy support in
>> Squid but will be applied to f
Hello,
This patch completes a series of Happy Eyeballs-related changes:
1. stable Squid: Parallel DNS A and queries.
2. v5 r15183: ASAP delivery of IPs from peer selection to FwdState.
3. This patch: ASAP delivery of IPs from DNS to peer selection.
4. A separate project should add:
On 06/21/2017 01:48 PM, Amos Jeffries wrote:
>> On 12.06.2017 14:14, Eduard Bagdasaryan wrote:
>>>
>>> On 01.06.2017 17:37, Amos Jeffries wrote:
The admin intention here is almost invariably to prevent certain
users getting cached data. Preventing cache being used simply for
lack of
On 06/23/2017 03:53 AM, Christos Tsantilas wrote:
> Στις 21/06/2017 08:07 μμ, ο Alex Rousskov έγραψε:
>> 2. *If* the request object is actually always there, then change the
>> pinConnection() parameter to an Http::Request reference (and change
>> callers to dereference t
On 06/21/2017 05:40 AM, Christos Tsantilas wrote:
> Replaced primary ConnStateData::pinConnection() calls with a pair of
> pinBusyConnection() and notePinnedConnectionBecameIdle() calls,
> depending on the pinned connection state ("busy" or "idle").
>
> Removed pinConnection() parameters that
On 06/21/2017 08:24 AM, Amos Jeffries wrote:
> On 21/06/17 23:40, Christos Tsantilas wrote:
>> * Protect Squid Client classes from new requests that compete with
>> ongoing pinned connection use and
>> * resume dealing with new requests when those Client classes are done
>> using the pinned
On 06/20/2017 08:43 AM, Amos Jeffries wrote:
> On 20/04/17 05:40, Amos Jeffries wrote:
>> On 20/04/17 04:26, Alex Rousskov wrote:
>>> On 03/26/2017 09:20 PM, Amos Jeffries wrote:
>>>> Any other issues that dont have bug reports I should wait for?
>>> I re
On 06/20/2017 03:39 AM, Amos Jeffries wrote:
> On 20/06/17 21:06, Christos Tsantilas wrote:
>> Στις 20/06/2017 01:59 πμ, ο Alex Rousskov έγραψε:
>>>> revno: 15212
>>>> committer: Amos Jeffries <squ...@treenet.co.nz>
>>>> branch nick: 5
>>
On 06/18/2017 07:39 AM, Amos Jeffries wrote:
> So, I update my vote to +1.
>
> Since you seem okay with this Alex I am assuming that is another +1 from
> you and applying to Squid-5.
I actually wanted to apply a couple of polishing touches before
committing. Now done in v5 r15210.
Thank you,
nstead of running with a WARNING,
but the patch does not change that behavior so we should not discuss it
here.
I hope my explanation of what the patch does will allow you to change
your vote.
Thank you,
Alex.
>> On 15.06.2017 17:35, Alex Rousskov wrote:
>>> On 06/15/2017 04
On 06/15/2017 04:57 AM, Andreas Weigel wrote:
> From discussion on squid-dev, the following behavior is implemented by
> this patch:
>
> * -N: The initial process is a master and a worker process.
> No kids.
> No daemonimization.
>
> * --foreground: The initial process is the master process.
On 06/09/2017 09:21 AM, Andreas Weigel wrote:
> Alex Rousskov wrote:
>> I would _not_ change that terminology now because
>> all the renaming will obfuscate your true fix and make it more difficult
>> to understand/review.
> I have to disagree to that one.
Renaming
On 06/09/2017 05:34 AM, Andreas Weigel wrote:
> - a SIGTERM to the master process makes it wait for its children to
> shutdown for 30 seconds
> --> Is that desired behavior?
The question is out of scope, but the answer is "yes". See
shutdown_lifetime in squid.conf.documented.
> I would
On 06/09/2017 01:38 AM, Andreas Weigel wrote:
> The
> overall idea was to make the decision to create and manage kids to
> depend on the number of configured workers/diskers.
It feels like you may be drifting into a dangerous territory of
fixing/changing "everything" related to SMP. Please stay
On 06/09/2017 07:23 AM, Amos Jeffries wrote:
> Since 3.5 is in stable it is quite limited in the types of
> things I'm accepting ports for. Changing the -N behaviour is not one of
> those due to the fragile net of things people have depending on it.
I do not think we should be changing -N
On 06/08/2017 09:36 PM, Stéphane Ouellette wrote:
> Isn't storeDirGetBlkSize() meant to return the "file system block size"?
Naturally, storeDirGetBlkSize() is not meant to return the file system
block size.
AFAICT, storeDirGetBlkSize() users expect this function to return,
roughly, the disk
On 06/08/2017 09:06 PM, Stéphane Ouellette wrote:
>while I was testing the use of the rock store on a ZFS pool using
> file system block sizes of 4K, 8K and 16K, I noticed that cachemgr.cgi
> always printed "FS Block Size = 1024" in all cases.
>
>I tracked down the issue by comparing
On 05/24/2017 10:27 PM, Amos Jeffries wrote:
> Okay. +1 on the code now
Committed to v5 (r15183). The next step is in the works.
Alex.
> 1. Squid currently implements: Use parallel A and queries.
> 2. This patch: ASAP delivery of IPs from peer selection to FwdState.
> 3. The next step:
On 06/08/2017 05:49 AM, Andreas Weigel wrote:
> Being interested in the SMP feature of squid we face the problem that it
> only works in daemon mode, which is inherently incompatible with runit
> supervising the service (would notice the process exiting and start it
> over and over again).
Hi
On 06/01/2017 06:38 PM, Alex Rousskov wrote:
> On 05/17/2017 06:20 PM, Alex Rousskov wrote:
>> Hello,
>>
>> While triaging a silent crash of a production Squid, I realized that
>> our death reporting code has a serious (and recently grown) hole. The
>> attache
On 05/20/2017 10:31 AM, Alex Rousskov wrote:
> On 05/20/2017 05:04 AM, Amos Jeffries wrote:
>
>> in src/Debug.h:
>>
>> * instead of adding yet another wrapper macro, please replace all uses
>> of debug_log with DebugStream().
> I will not do that because ma
On 05/17/2017 06:20 PM, Alex Rousskov wrote:
> Hello,
>
> While triaging a silent crash of a production Squid, I realized that
> our death reporting code has a serious (and recently grown) hole. The
> attached patch closes it. This reporting-only patch does not fi
On 05/22/2017 10:26 AM, Alex Rousskov wrote:
> On 05/08/2017 08:03 AM, Eduard Bagdasaryan wrote:
>> This patch fixes [reopened] bug 2833.
>>
>> A security fix made in r14979 had a negative effect on collapsed
>> forwarding. All "private" entries were considered
On 05/30/2017 10:58 PM, Amos Jeffries wrote:
> On 26/05/17 22:08, Christos Tsantilas wrote:
>> --consensus allows matching a part of the conglomerate when the part's
>> subject name is included in certificates used by many other
>> conglomerate parts (e.g., matching Google but not Youtube).
> So
If there are none, file a new bug report.
Alex
> - Original Message -----
> From: "Alex Rousskov" <rouss...@measurement-factory.com>
> To: "William Lima" <william.l...@hscbrasil.com.br>
> Cc: squid-dev@lists.squid-cache.org
> Sent: Wednesday, May 24,
On 05/24/2017 09:04 AM, Amos Jeffries wrote:
> For the 1st destination ps_state::handlePath() calls noteDestination()
> which results in the entire TunnelState::startConnecting() /
> FwdState::startConnectionOrFail() sequence happening synchronously
> before returning to ps_state::handlePaths()
On 05/24/2017 11:29 AM, William Lima wrote:
>
> Sharing a single cache_dir (diskd) entry for a multiple worker setup is a
> valid approach, right?
No, it is not valid. Diskd is not SMP-aware. Using a single diskd
cache_dir with multiple workers results in undefined behavior which may
include
On 05/24/2017 02:31 AM, Christos Tsantilas wrote:
> Adds support for --long-acl-options
s/Adds/Add/ or
s/Adds support for/Support/.
> This patch adds support for --long-acl-options. The old single-letter
The first sentence in the second paragraph can be removed as it
duplicates the first
On 05/15/2017 11:08 AM, Alex Rousskov wrote:
> On 05/15/2017 08:31 AM, Eduard Bagdasaryan wrote:
>> I followed your plan with few adjustments and reattached the patch.
>
> LGTM. Will commit soon if there are not objections.
Committed to v5 (r15145) with minor polishing touches.
T
On 04/29/2017 10:17 AM, Alex Rousskov wrote:
> On 04/29/2017 02:33 AM, Amos Jeffries wrote:
>> +1 if you update the patch to do the above. +0 otherwise - ie. go ahead
>> with commit if you want, but I'd rather see the fixed caller(s) fixed to
>> use a more long-ish term f
On 05/08/2017 08:03 AM, Eduard Bagdasaryan wrote:
> This patch fixes [reopened] bug 2833.
>
> A security fix made in r14979 had a negative effect on collapsed
> forwarding. All "private" entries were considered automatically
> non-shareable among collapsed clients. However this is not true: there
On 05/20/2017 05:04 AM, Amos Jeffries wrote:
> in src/Debug.h:
>
> * instead of adding yet another wrapper macro, please replace all uses
> of debug_log with DebugStream().
I will not do that because many (most?) debug_log users should not use
DebugStream(). They should use the standard
Hello,
The attached patch allows Squid to report various problems (e.g.,
mishandled exceptions) that may occur very early in Squid lifetime,
before proper logging is configured by the first _db_init() call.
To enable such early reporting, we started with a trivial debug.cc change:
-FILE
Amos Jeffries wrote:
> Please use std::atomic
> locks to ensure actually atomic operations on the file locks.
std::atomic<> is defined for integers and pointers. It is not defined
for "file locks", which are complex OS-maintained objects largely
invisible to Squid. C++11 std::atomic cannot
On 05/18/2017 10:23 AM, Christos Tsantilas wrote:
> On 18/05/2017 06:05 μμ, Alex Rousskov wrote:
>> I suspect it would be cheaper, in the long term, to use feature tests.
> However is not easy to always use this method. The OpenSSL-1.1.0 put
> many API changes trying to hide st
On 05/18/2017 09:34 AM, Ingo Schwarze wrote:
> Alex Rousskov wrote on Thu, May 18, 2017 at 09:05:29AM -0600:
>> On 05/18/2017 05:12 AM, Christos Tsantilas wrote:
>>> Agrr... Using the openSSL version was the faster/easier way. Touching
>>> autoconf may result
On 05/18/2017 05:12 AM, Christos Tsantilas wrote:
> Agrr... Using the openSSL version was the faster/easier way. Touching
> autoconf may result to 2-3 full squid rebuilds to implement/test similar
> fixes.
The alternative is to convince others that Squid will not support
OpenSSL API
Hello,
While triaging a silent crash of a production Squid, I realized that
our death reporting code has a serious (and recently grown) hole. The
attached patch closes it. This reporting-only patch does not fix any
crashes. There are more technical details in the preamble.
Researching this
On 05/17/2017 03:18 PM, David Hogan wrote:
> I found that applying a blacklist at step3 resulted in too many false
> positives
> caused by subjectAltName matches.
Factory is working on a patch to address that problem.
> I am hoping separately to figure
> out how to match missing SNI and
On 05/17/2017 12:09 PM, David Hogan wrote:
> ssl_bump peek step1
> ssl_bump peek step2 whitelist
> ssl_bump terminate step2 blacklist
> ssl_bump peek step2
> ssl_bump splice step3
The above configuration looks strange but this is squid-dev not
squid-users, so I trust you
On 05/17/2017 10:35 AM, Christos Tsantilas wrote:
> +#if (OPENSSL_VERSION_NUMBER >= 0x10002000L)
> +X509 * cert = SSL_CTX_get0_certificate(ctx.get());
If it is possible to replace this version check with a ./configure-time
detection of SSL_CTX_get0_certificate() availability, please do that.
On 05/16/2017 03:31 AM, Amos Jeffries wrote:
> On 16/05/17 16:33, Alex Rousskov wrote:
>> The attached patch is the first in a short series of patches that
>> improve Squid support for the Happy Eyeballs principle. The motivation
>> for these patches has been discussed on
On 05/16/2017 06:23 AM, Amos Jeffries wrote:
> On 16/05/17 06:53, Eduard Bagdasaryan wrote:
>> Ssl::Lock class resides in
>> src/security/cert_generators/file/certificate_db.h.
> I suspect the limitations you found to that class are behind the
> ssl_crtd helper bugs
Those Ssl::Lock limitations
On 05/15/2017 08:31 AM, Eduard Bagdasaryan wrote:
> I followed your plan with few adjustments and reattached the patch.
LGTM. Will commit soon if there are not objections.
Alex.
> On 06.05.2017 00:33, Alex Rousskov wrote:
>> On 04/27/2017 02:39 PM, Eduard Bagdasaryan wrote:
>&g
On 03/31/2017 07:21 AM, Christos Tsantilas wrote:
> Avoid sending second CONNECT request to adaptation
>
>
> The users may not want to send the second request to the adaptation
> services. In this case they can use acls as follows:
>
> acl
On 04/27/2017 02:39 PM, Eduard Bagdasaryan wrote:
> +// always start probing in order to effectively detect
> +// dead or revived peers
> +(void)peerProbeConnect(p);
I think we can simplify that comment while making it more precise:
peerProbeConnect(p); // detect any died or
On 05/01/2017 07:38 AM, Eduard Bagdasaryan wrote:
> There are still no remarks about this patch.
> Is there something to be fixed before it can go in?
Committed to v5 (r15133) without changes.
Thank you,
Alex.
> On 18.04.2017 13:40, Eduard Bagdasaryan wrote:
>> Hello,
>>
>> This patch removes
no reason to resist supporting "has ALE" if Squid support for ALE
is currently broken and cannot be fixed easily/quickly. If you see flaws
in this logic, please point them out. I may be missing something.
Thank you,
Alex.
> On 01.05.2017 19:11, Alex Rousskov wrote:
>> On 04/30/201
On 04/30/2017 11:26 PM, Amos Jeffries wrote:
> On 30/04/17 04:17, Alex Rousskov wrote:
>> On 04/29/2017 02:33 AM, Amos Jeffries wrote:
>>
>>> If that is too hard, making callers do the allocate bit themselves and
>>> xstrncpy() instead should be a good medi
On 04/30/2017 10:03 PM, Amos Jeffries wrote:
> Is there an explicit need you have found for ALE to be on the
> component list? Since ALE is currently standing in as a "master
> transaction" object for most of the Squid code. It needs to be either
> created or provided/fetched from elsewhere
On 04/29/2017 02:45 AM, Amos Jeffries wrote:
>> "auto" needs "const" and "&" but not "*".
> FYI: that statement does not match my experiences with auto in the
> general case.
I suspect it matches your experiences but not your interpretation of
those experiences. Squid has many examples where
Hello,
I accidentally discovered that our xstrndup() documentation lies and
its implementation is dangerously nothing like the standard one. The
attached patch fixes documentation and three callers. Please see the
patch preamble for details, including lack of ESI testing.
The attached patch
On 04/26/2017 06:53 AM, Amos Jeffries wrote:
> Been digging around in library guts and yes you were right Alex - for
> OpenSSL and the very old ones based on SSL protocol. Just not for the
> other N libraries based on TLS design which I/we hope to support. There
> is structural difference between
On 04/19/2017 08:06 AM, Loic Dachary wrote:
> On 04/19/2017 03:53 PM, Alex Rousskov wrote:
>> On 04/18/2017 01:00 AM, Loic Dachary wrote:
>>
>>> It turns out [CARP] performs as well as CRUSH
>> On 04/19/2017 06:51 AM, Loic Dachary wrote:
>>
>>
On 04/17/2017 10:53 AM, Loic Dachary wrote:
> On 04/17/2017 06:28 PM, Alex Rousskov wrote:
>> On 04/17/2017 09:08 AM, Loic Dachary wrote:
>>
>>> peer 1 targeted by a set of requests (X1)
>>> peer 2 targeted by a set of requests (X2)
>>
>>&
On 04/17/2017 09:08 AM, Loic Dachary wrote:
> peer 1 targeted by a set of requests (X1)
> peer 2 targeted by a set of requests (X2)
> and we add a new peer, CRUSH makes it so 1/3 of the requests from peer 1 and
> 2 move to peer 3
> peer 1 targeted by a set of requests (2/3 of X1)
>
On 04/12/2017 10:13 AM, Amos Jeffries wrote:
> The current O(1) is achieved only through the way dlink requires an
> iterator raw-pointer to be stored in the Entry itself, not the list
> owner object.
Yes, dlink is an intrusive list.
> That itself is a yuck factor that causes us a lot of
>
On 04/12/2017 09:53 AM, Amos Jeffries wrote:
> I have been thinking of this again and think I have come up with a way
> to make some better progress.
>
> The idea is to split the NULL -> nullptr changes into two groups:
>
> 1) those lines where sed s/nullptr/NULL/g can be used to reverse the
>
On 04/12/2017 09:30 AM, Adam Majer wrote:
>>> If you need an intrusive list container for performance reasons, please
>>> consider using Boost.Intrusive instead of rolling your own.
>> The idea of using Boost for Squid has been discussed and, IIRC, rejected
>> (for the time being): We have more
On 04/12/2017 06:08 AM, Adam Majer wrote:
> On 04/12/2017 03:48 AM, Alex Rousskov wrote:
>> BTW, if you do end up removing this intrusive list, please check whether
>> its ExternalACLEntry::lru "anchor" member should be removed as a side
>> effect. P
On 03/27/2017 12:35 AM, 钱国正 wrote:
> What if the protocol is not http, but using 80 port, can squid relay it
> to its target server?
Yes, if you intercept port 80 connections. Assuming you do intercept,
what you want should be already supported by on_unsupported_protocol.
However, there is a
On 03/24/2017 04:54 AM, 钱国正 wrote:
> I want to know what's the
> pinning.serverConnection mean? and what it is used for?
Before we dive into low-level details, please allow me to ask an
important high-level question. Your answer may render those low-level
detail irrelevant:
> I want to rewrite
On 03/22/2017 07:53 PM, Amos Jeffries wrote:
> This helpers' passive mode does not need to care what the input is, so
> long as it is consistent.
I do not recommend ignoring input format in hope that ignored input is
consistent/acceptable/compatible/etc. However, I am not going to object
to any
On 03/22/2017 06:44 AM, Amos Jeffries wrote:
> This is a side effect of the change to using logformat tokens. The
> %DATA field is always present now.
My understanding is that, after the logformat changes, Squid started
appending "-" to the helper query when an external ACL has no
parameters.
On 03/21/2017 10:27 PM, Amos Jeffries wrote:
> On 22/03/2017 10:27 a.m., Alex Rousskov wrote:
>> This Request For Comments proposes to remove a subtle Squid
>> (mis)feature. If you happen to use the corresponding feature, please
>> speak up to protect it! If nobody defends
On 03/22/2017 07:20 AM, Amos Jeffries wrote:
> On 17/03/2017 6:17 a.m., Alex Rousskov wrote:
>> On 03/16/2017 05:15 AM, Amos Jeffries wrote:
>>
>>
>>> Any objections to applying this with this added:
>>>
>>> // XXX: putStr() still has String
On 03/17/2017 09:48 AM, Heiler Bemerguy wrote:
> Sadly the same thing occurred again. It seems the hole is deeper lol..
Most likely, it is the same hole. However, the more we panic and jump to
conclusions, the deeper that hole below us may look...
> I couldn't find any previous messages that
On 03/17/2017 02:05 AM, Gary Wang wrote:
> I made another change to keep new DEFAULT_IPC_PREFIX
> consistent in src/ipc/Port.cc file.
>Could you guys help review the MP?
>https://code.launchpad.net/~gary-wzl77/squid/ipc_prefix/+merge/318714
I can only repeat what I have already
On 03/17/2017 01:06 AM, 钱国正 wrote:
> I am using squid as a transparent proxy in Router,
> redirecting data which destination port is 80 to squid,
> and it work fine for a long time, just days before, I found there is one
> case that will not make the mobile app not work.
>
> The case is the
On 03/16/2017 05:30 PM, Gary Wang wrote:
> On Thu, Mar 16, 2017 at 2:49 AM, Alex Rousskov wrote:
>> On 03/15/2017 11:03 AM, Gary Wang wrote:
>>>In snap world, only the following name pattern is allowed
>>> /dev/shm/sem.snap.@{SNAP_NAME}.*
>>>
On 03/16/2017 11:00 AM, khaled belhout wrote:
> but we need to annotate the some functions with [[noreturn]] attribute like :
I do not think such annotations belong to the source code. If we add
various annotations for various tools, the code readability and
maintainability will suffer.
AFAIK,
On 03/16/2017 01:15 AM, Amos Jeffries wrote:
> * "start" is an action name and we use it (almost?) exclusively for Job
> initiation. By comparison "started" means/implies more clearly a state
> or time point. The Tunnel member stores a time point.
> - IMO both are bad, but "started" is better
On 03/16/2017 05:15 AM, Amos Jeffries wrote:
> Any objections to applying this with this added:
>
> // XXX: putStr() still has String 64KB limits
> Must(strVia.length() < 64*1024);
No objections from me if you replace the magic constant with a new
inlined String::MaxSizeXXX() method. The
On 03/16/2017 01:43 AM, Amos Jeffries wrote:
> On 6/03/2017 3:54 p.m., Alex Rousskov wrote:
>> On 03/05/2017 12:28 PM, Eduard Bagdasaryan wrote:
>>> ufsdump build is broken now [...]
>>> This patch fixes this, however I am not sure that
>>> does it in a best po
On 03/15/2017 11:03 AM, Gary Wang wrote:
> On Thu, Mar 16, 2017 at 12:33 AM, Alex Rousskov wrote:
>> On 03/15/2017 03:24 AM, Gary Wang wrote:
>> > Regarding the confinement of usage of shared memory in snap world,
>> >Please take a look at th
On 03/15/2017 08:17 AM, Amos Jeffries wrote:
> That said, there are points in the lifecycle which are more friendly to
> bg changes. The next one I expect these C++11 changes can even have a
> chance is after 3.5 ceases to be a supported version.
I agree, especially if "supported" includes "we
acement noise.
In summary: I still do not think these changes are desirable (even in
their polished variant), but if others overrule me on that, my comments
about the quality of the current changes still apply.
Thank you,
Alex.
> 2017-03-13 14:45 GMT+01:00 Alex Rousskov <rouss...@measur
On 03/14/2017 09:52 PM, Gary Wang wrote:
> On Tue, Mar 14, 2017 at 11:49 PM, Alex Rousskov wrote:
>> On 03/14/2017 08:44 AM, Gary Wang wrote:
>> > About the DEFAULT_STATEDIR,
>> > ...
>> > DEFS += -DDEFAULT_S
On 03/14/2017 10:43 AM, Heiler Bemerguy wrote:
> Em 07/03/2017 20:26, Alex Rousskov escreveu:
>> How can a disker response get stuck? Most likely, something unusual
>> happened ~13 days ago. This could be a Squid bug and/or a kid restart.
> root@proxy:~# ps auxw |grep squid
On 03/14/2017 08:44 AM, Gary Wang wrote:
> About the DEFAULT_STATEDIR,
> ...
> DEFS += -DDEFAULT_STATEDIR=\"$(localstatedir)/run/squid\"
> ...
> According to the above share memory namespace in snap world, it
> couldn't help on this either.
Here is how I [mis]interpret the
On 03/13/2017 08:25 AM, Eduard Bagdasaryan wrote:
> On 14.02.2017 04:22, Amos Jeffries wrote:
>> The problem is with proxy where the admin has configured large headers
>> to be allowed, and receives a Via just under the 6KB liit. Our append
>> pushing it over by even one byte would assert.
Yes,
On 03/13/2017 09:38 AM, Heiler Bemerguy wrote:
>
> root@proxy:/var/log/squid# cat cache.log |grep assertion
> 2017/03/13 07:50:54 kid6| assertion failed: client_side_reply.cc:1167:
> "http->storeEntry()->objectLen() >= headers_sz"
> 2017/03/13 08:17:46 kid4| assertion failed:
are good and new code should use them!
However, I doubt the advantages of changing those old loops outweigh
cross-branch development costs right now. Others may disagree, and, if
they do, I would not object to a polished patch being committed.
Thank you,
Alex.
> 2017-03-12 16:31 GMT+01:00 Al
On 03/12/2017 07:45 AM, khaled belhout wrote:
> this patch modernize for loops using c++11 Range-based for loop
Please use "const auto" where possible and avoid using "i" for naming
the current iteration object.
I am curious why did you decide to change all these loops? How did you
select the
On 03/09/2017 06:51 PM, Amos Jeffries wrote:
> On 6/03/2017 6:02 a.m., Eduard Bagdasaryan wrote:
>> This patch improves Squid to better distinguish error cases when loading
>> cache entry metadata is failed. Knowing the exact failure reason may
>> help triage and guide development. Refactoring
On 03/07/2017 01:08 PM, Heiler Bemerguy wrote:
> Some log from right now...
Here is my analysis:
> 15:53:05.255| ipc/Queue.h findOldest: peeking from 7 to 6 at 1
Squid worker (kid6) is looking at the queue of disker (kid7) responses.
There is just one response in the queue.
> IpcIoFile.cc
On 03/07/2017 10:58 AM, Heiler Bemerguy wrote:
> I used iostat to check if "right now" the hds were being accessed. A
> lot of minutes passed and all writes/reads remained Zero.
Understood.
> With a 80mbit/s traffic going on, how could nothing be written nor read from
> disc?
I can come up
201 - 300 of 788 matches
Mail list logo