OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.0 - page 564.

crazyates

legendary

Activity: 952

Merit: 1000

Quote from: Hyphen on June 01, 2012, 03:24:43 PM

Quote from: crazyates on June 01, 2012, 03:22:30 PM

Quote from: Hyphen on June 01, 2012, 03:16:14 PM

For some reason, if cgminer crashes badly ( not my computer, my computer is still up and running. )
it appears that it deletes the cgminer executable.
All the other files in the cgminer folder are there, just not the executable.
Anybody else experience this?
Has been happening to me about once every 3 weeks since around cgminer 2.3.3 I think.
I'm running windows 7 and currently using cgminer 2.4.1, just experienced this issue again.
Not sure what could be causing this. It's not standard hard disk data corruption, it's happened under 3 computers of mine now and no other programs seem to be effected.

Anti-virus? Some anti-virus programs just delete "suspected files" without any user prompt.

I'm using avast. Pretty sure avast would just sandbox it, and then notify me of that.

*edit* It does appear to be avast. It started deleting it every time I copied over on one computer until I disabled avast.

Anyway, if anybody else is using avast and experiences this problem, well, here's your explanation.

Weird though... Usually avast doesnt act like this.

Trend Micro does the same thing - just deletes the .exe. I switched to MSE and it wants to get rid of it, but I told it to ignore.

af_newbie

legendary

Activity: 2702

Merit: 1468

Question for Con or Kano?

Should GPU statistics be reset when a GPU is disabled?

What I see is that "avg", "util", falls off slowly when GPU is disabled.
After some time the avg, util values are very small and when the GPU is re-enabled,
the statistics is not correct for sometime as the duration of the OFF time is taken into account.

~~Maybe the GPU stats should be reset when GPU is disabled.~~

EDIT: Maybe total_secs should take into account (subtract) the time spent in disabled state.
This should "freeze" all time sensitive stats.

Hyphen

newbie

Activity: 20

Merit: 0

Quote from: crazyates on June 01, 2012, 03:22:30 PM

Quote from: Hyphen on June 01, 2012, 03:16:14 PM

For some reason, if cgminer crashes badly ( not my computer, my computer is still up and running. )
it appears that it deletes the cgminer executable.
All the other files in the cgminer folder are there, just not the executable.
Anybody else experience this?
Has been happening to me about once every 3 weeks since around cgminer 2.3.3 I think.
I'm running windows 7 and currently using cgminer 2.4.1, just experienced this issue again.
Not sure what could be causing this. It's not standard hard disk data corruption, it's happened under 3 computers of mine now and no other programs seem to be effected.

Anti-virus? Some anti-virus programs just delete "suspected files" without any user prompt.

I'm using avast. Pretty sure avast would just sandbox it, and then notify me of that.

*edit* It does appear to be avast. It started deleting it every time I copied over on one computer until I disabled avast.

Anyway, if anybody else is using avast and experiences this problem, well, here's your explanation.

Weird though... Usually avast doesnt act like this.

crazyates

legendary

Activity: 952

Merit: 1000

Quote from: Hyphen on June 01, 2012, 03:16:14 PM

For some reason, if cgminer crashes badly ( not my computer, my computer is still up and running. )
it appears that it deletes the cgminer executable.
All the other files in the cgminer folder are there, just not the executable.
Anybody else experience this?
Has been happening to me about once every 3 weeks since around cgminer 2.3.3 I think.
I'm running windows 7 and currently using cgminer 2.4.1, just experienced this issue again.
Not sure what could be causing this. It's not standard hard disk data corruption, it's happened under 3 computers of mine now and no other programs seem to be effected.

Anti-virus? Some anti-virus programs just delete "suspected files" without any user prompt.

Hyphen

newbie

Activity: 20

Merit: 0

For some reason, if cgminer crashes badly ( not my computer, my computer is still up and running. )
it appears that it deletes the cgminer executable.
All the other files in the cgminer folder are there, just not the executable.
Anybody else experience this?
Has been happening to me about once every 3 weeks since around cgminer 2.3.3 I think.
I'm running windows 7 and currently using cgminer 2.4.1, just experienced this issue again.
Not sure what could be causing this. It's not standard hard disk data corruption, it's happened under 3 computers of mine now and no other programs seem to be effected.

kano

legendary

Activity: 4634

Merit: 1851

Linux since 1997 RedHat 4

Quote from: kano on May 31, 2012, 12:39:45 AM

Quote from: dmcurser on May 30, 2012, 06:06:00 PM

everytime i try to run java api notify i get a constant tag 13 in class file

Well "code tag" the actual error so I can see.
Also include the output of "java -version"
and lastly - of course it is "java API notify"
(the capital letters I've typed before do matter)
(though that is probably nothing to do with the problem)

The fix may well be "javac API.java" but will require the java SDK

Was helping someone in IRC who got the same error.
I recompiled API.java (with an old java 6.0_03 SE) and got him to download it and it worked.
Only to later find that the files were actually the same - the issue was the file protection.
I've sent a pull to the git with the new API.class it only shows a file protection difference.
So you may be able to fix it by changing the file protection of API.class in windows to be executable - though I'm really not sure of the standard method to do that in windows

zefir

donator

Activity: 919

Merit: 1000

Quote from: kano on June 01, 2012, 05:10:26 AM

The Icarus code was fixed by xiangfu 3 weeks ago that seems to match your exact change you mentioned on the previous page so I guess that's where you got that code from?
https://github.com/ckolivas/cgminer/commit/b1785d1f07d9482b5ed6104c87b39a3888fa8efb

BFL's - well - I'm not sure where the BFL code is going at the moment ...

Yes, I basically compared the difference between drivers and found that only the bitforce was missing the additional check for MIBSEB builds.

Not sure how to interpret your last sentence, but it would be a pity if the support for some devices got negatively influenced by political dissents Sad

zefir

donator

Activity: 919

Merit: 1000

Quote from: -ck on June 01, 2012, 04:38:30 AM

Quote from: zefir on June 01, 2012, 01:28:45 AM

Any reason not to use ntohl()/htonl() without explicit checking for endianess and let the the preprocessor taking care on that?

No

It is not a big issue and I'd have already submitted the related patch, but I can't test Windows builds ATM and refrained from posting untested code...

kano

legendary

Activity: 4634

Merit: 1851

Linux since 1997 RedHat 4

Quote from: zefir on June 01, 2012, 01:28:45 AM

Quote from: -ck on May 31, 2012, 04:40:18 PM

It's fixed in the git tree so grab a tarball instead from git or use git to grab the latest code.

You mean the endian thingy for BFLS? Not sure, the condition in driver-bitforce.c@HEAD:357 is still checked against __BIG_ENDIAN__, which seems to be not set in the gcc-Linaro-4.6 toolchain for building MIPS.

...

The Icarus code was fixed by xiangfu 3 weeks ago that seems to match your exact change you mentioned on the previous page so I guess that's where you got that code from?
https://github.com/ckolivas/cgminer/commit/b1785d1f07d9482b5ed6104c87b39a3888fa8efb

BFL's - well - I'm not sure where the BFL code is going at the moment ...

-ck

legendary

Activity: 4088

Merit: 1631

Ruu \o/

Quote from: zefir on June 01, 2012, 01:28:45 AM

Any reason not to use ntohl()/htonl() without explicit checking for endianess and let the the preprocessor taking care on that?

No

zefir

donator

Activity: 919

Merit: 1000

Quote from: -ck on May 31, 2012, 04:40:18 PM

It's fixed in the git tree so grab a tarball instead from git or use git to grab the latest code.

You mean the endian thingy for BFLS? Not sure, the condition in driver-bitforce.c@HEAD:357 is still checked against __BIG_ENDIAN__, which seems to be not set in the gcc-Linaro-4.6 toolchain for building MIPS.

Any reason not to use ntohl()/htonl() without explicit checking for endianess and let the the preprocessor taking care on that?

kano

legendary

Activity: 4634

Merit: 1851

Linux since 1997 RedHat 4

Quote from: Un zafado cualquiera on May 31, 2012, 11:01:14 PM

Quote from: kano on May 31, 2012, 10:53:05 PM

Compile it yourself - it's an option

I was trying not being a jerk. I think is more useful for the newbies not givin anything suspicius. Just in case its end bad.

Ah you mean to have it always like that.
No - I can't see that ever happening.

Though I would point out the obvious: you're solution would not fix the problem.
They would simply identify something else in cgminer to tag it as a virus.

Windows virus checkers are inherently crap in this case, since they identify part of cgminer rather than the virus that installed cgminer.

Of course if someone wrote a virus that installed the anti-virus program ... of course they would fix that problem ... but in this case I guess they are just a bunch of loser programmers who can't see the blatantly obvious crappyness of what they are doing in this case.

Un zafado cualquiera

full member

Activity: 158

Merit: 100

aquí dice algo personal.

Quote from: kano on May 31, 2012, 10:53:05 PM

Compile it yourself - it's an option

I was trying not being a jerk. I think is more useful for the newbies not givin anything suspicius. Just in case its end bad.

kano

legendary

Activity: 4634

Merit: 1851

Linux since 1997 RedHat 4

Compile it yourself - it's an option

Un zafado cualquiera

full member

Activity: 158

Merit: 100

aquí dice algo personal.

Is there a way to eliminate dependence on curses. There where a lot of antivirus reporting it as a keylogger and is probably a never ending cycle telling antivirus company to check correctly. Evry antivirus is reporting libpdcurses as a trojan.

-ck

legendary

Activity: 4088

Merit: 1631

Ruu \o/

It's fixed in the git tree so grab a tarball instead from git or use git to grab the latest code.

zefir

donator

Activity: 919

Merit: 1000

Quote from: P_Shep on May 31, 2012, 03:30:26 PM

Would CFLAG -D_BIG_ENDIAN_ not be an option?

Not if you want to remain endian agnostic, i.e. if you add -D_BIG_ENDIAN_ to CFLAGS in the OpenWRT Makefile, you'll need to remove it for little endian builds.

For Linux builds the way it is done in cgminer is wrong, but I guess it is so to compile on Windows. Usually you add bswapXX unconditionally to any network byte order (or big endian) value and the preprocessor resembles that call to an empty macro on big endian machines. In fact, building cgminer for OpenWRT warns about multiple definitions of bswap32 that are already defined in .

There is some potential to fix this, but since I can not test Windows builds I'd take it as is with the given workaround.

P_Shep

legendary

Activity: 1795

Merit: 1208

This is not OK.

Quote from: zefir on May 31, 2012, 03:17:52 PM

Never mind, found the problem in __BIG_ENDIAN__ not being set for that target platform.

This patch is required to build cgminer with BFLS support for OpenWRT:

Code:

diff --git a/driver-bitforce.c b/driver-bitforce.c
index a59338c..d95f0eb 100644
--- a/driver-bitforce.c
+++ b/driver-bitforce.c
@@ -354,7 +354,7 @@ static uint64_t bitforce_scanhash(struct thr_info *thr, struct work *work, uint6
 
 	while (1) {
 		hex2bin((void*)&nonce, pnoncebuf, 4);
-#ifndef __BIG_ENDIAN__
+#if !defined (__BIG_ENDIAN__) && !defined(MIPSEB)
 		nonce = swab32(nonce);
 #endif

Good Luck!

Would CFLAG -D_BIG_ENDIAN_ not be an option?

zefir

donator

Activity: 919

Merit: 1000

Never mind, found the problem in __BIG_ENDIAN__ not being set for that target platform.

This patch is required to build cgminer with BFLS support for OpenWRT:

Code:

diff --git a/driver-bitforce.c b/driver-bitforce.c
index a59338c..d95f0eb 100644
--- a/driver-bitforce.c
+++ b/driver-bitforce.c
@@ -354,7 +354,7 @@ static uint64_t bitforce_scanhash(struct thr_info *thr, struct work *work, uint6
 
 	while (1) {
 		hex2bin((void*)&nonce, pnoncebuf, 4);
-#ifndef __BIG_ENDIAN__
+#if !defined (__BIG_ENDIAN__) && !defined(MIPSEB)
 		nonce = swab32(nonce);
 #endif

Good Luck!

zefir

donator

Activity: 919

Merit: 1000

Anyone being successful operating BFLS with cgminer from a big endian machine?

I am using the same device xiangfu does for operating his Icarus farm (TP-Link TL-MR3020), which is based on an Atheros MIPS core (big endian). Using the git HEAD this is the debug output:

Code:

[2012-05-31 18:45:24] BitForce block data: a4870512b33f467106a9907295eec4a7031e2149e6b90d81d6efbee03d8c7302eb755e1c4fc7bc401a0a8b5f
[2012-05-31 18:45:29] BitForce waited 5200ms until NONCE-FOUND:01A85E01,8AD1912A
[2012-05-31 18:45:29] Proof: 570e5ed30f82f60447dcf34a4f00f733bb05e1d2b0f1120e54e9817faffa2f84
                      Target: 00000000ffffffffffffffffffffffffffffffffffffffffffffffffffffffff
                      TrgVal? no (false positive; hash > target)
[2012-05-31 18:45:29] Share below target
[2012-05-31 18:45:29] Proof: 1acf5c5b68608bf3858514064f899f439b030e30c5502c8e6020165414cadab7
                      Target: 00000000ffffffffffffffffffffffffffffffffffffffffffffffffffffffff
                      TrgVal? no (false positive; hash > target)
[2012-05-31 18:45:29] Share below target
[2012-05-31 18:45:29] [thread 0: 4294967295 hashes, 812342882 khash/sec]

I suspect something is not fully endian aware in the driver-bitforce module. I compared to what xiangfu/kano did for Icarus, but the relevant code lines seem to match the bitforce counterpart.

Any idea where to look at first?

Thanks, zefir

Topic: OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.0 - page 564. (Read 5806088 times)