Hi Igor,
I re-started to recover my 73 GiB archive created earlier with p7zip under Linux now with 7-Zip 21.02 alpha under Win 10. I pushed from the drop down menu 'Open inside #' item and I got 52 different zip files now. (se attached) I can see the offsets and the last zip may contains my zipped stuff.
Can you please provide me with an idea what shall I do to unpack the complete package? Thanks.
Or is there any tricks to concatenate zip file parts and unzip the whole stuff?
Well, I tried with Raspberry Pi4 @2 GHz for armhf(32bit), same settings as with 64-bit: https://sourceforge.net/p/sevenzip/discussion/45797/thread/d401ab2966/#b089/96da
but it crashed at crc32:32 4t every time. 1t and 2t worked fine, so it may has something to do with overclocking, so here are results for standard clocks with raspian 2021.1 for pi4 4GB:
The hardware instruction crc32 is very fast on Cortex-A72. Probably it can be executed with latency=1 on A72.
And Cortex-A57 shows 3 cycles latency for same instruction.
So A72 is 3 times faster there. All another things look similar for performance of A72 and A57, but A72 in Pi4 doesn't support SHA and AES instructions.
That CRC32 test is pretty fast on A72 to show the bandwidth of RAM and cache when the block is large.
You can run it on arm64 so:
7zz b -mm=CRC32:64 -mtime=2
Maybe you can use similar CRC32 benchmark also to catch exact lines with crash. You can change the time with -mtime=2 for seconds or with -mtimems=100 for ms, if you want faster runs.
And what is that crash on 2 GHz?
Is it some error message from 7-Zip program or it's some crash on system level?
Last edit: Igor Pavlov 2021-03-10
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
The crash was black screen and what appears to be an instant shutdown of SOC.
Strange, because on ARM64 it worked stable so far oc 2GHz, but not for this 32bit code.
My pi has heatsink and fan attached, and a very powerful PSU (not usb-c), so this was not the problem. It seems, the problem was GPU OC setting of 650 MHz, i changed that to 550, maybe that has caused a deadlock before (GPU doesn't do anything for 7zz, but on pi it also has the memory controller). Now it's working on 2GHz cpu clock.
On Ubuntu20.10 for Raspberrypi ARM64, you can add architecture armhf to dpkg and to repositories, and install the required libraries, libc6 and libstdc++ to get the 32bit armhf binary working. https://wiki.debian.org/Multiarch/HOWTO
I have added below raspberrypi4 armhf and the ARM64 benchmarks from the other thread for comparison.
I installed multiarch armhf + ARM64 on ubuntu 20.10 ARM64 for pi4, and was able to run 32 Bit code there too, now even with overclock settings (which crashed on raspbian 32Bit):
Still i think GPU clock 650 is considered to be unstable, since it crashed before on raspian.
Something else is different. On raspian there was 32Bit kernel.
I have been using 7-Zip for 64-bit Linux x86-64 (AMD64) since the first day of this version's release. I replaced the jinfeihan57/p7zip fork in the Amavis configuration on my test server without any problems, nothing recorded in the logs.
+1 for adding Zstandard.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Don't know if this is intentional, but the -o switch works in a funny way. Let's say I downloaded the source code for 7z and wanted to extract it in a directory same as the filename of the archive. Adding a space between -o and the output directory breaks the switch.
$ ~/.7z2101-linux-x64/7zz x 7z1900-src.7z -o ./7z1900-src
Command Line Error:
Too short switch:
-o
$ ~/.7z2101-linux-x64/7zz x 7z1900-src.7z -o./7z1900-src
7-Zip (z)21.01 alpha (x64) : Copyright (c)1999-2021 Igor Pavlov : 2021-03-09
compiler: 9.3.0 GCC 9.3.0 64-bit locale=en_US.UTF-8 Utf16=on HugeFiles=on CPUs:4 Intel(R) Core(TM) i7-5550U CPU @ 2.00GHz (306D4),ASM,AES
Scanning the drive for archives:
1 file, 1071569 bytes (1047 KiB)
Extracting archive: 7z1900-src.7z
--
Path= 7z1900-src.7z
Type= 7z
Physical Size=1071569
Headers Size=19163Method= LZMA:23
Solid= +
Blocks=1
Everything is Ok
Files: 1172
Size: 7339189
Compressed: 1071569
Just wanted to drop another note of thanks for the native Linux support. Very welcome and surprising news, especially considering p7zip seems to be abandoned at this point.
Last edit: Jared B. 2021-03-11
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I've just checked it on Fedora 33 and it worked fine. However, I've noticed that with -p it prints a given password while typing. Is there a switch to hide it?
In addition to that, for -p there was a nice feature in p7zip to ask twice for manually typed (and hidden) password. Thanks to that you didn't need to write the password once and then test the archive to verify that you entered the right password. Typing it twice was very effective in my case to eliminate typos. IT would be nice to have it also available in the new version.
❤️
1
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I hope he has reached out to you instead of just posting and claiming to not want to post it publicly.
Considering you don't normally release source until its final is this scare mongering (if he hasn't reached out)?
He has since gone on to "elaborate" on his "strongly" worded first post, naturally he didn't understand the power of words with regards to the OBVIOUS worry it would cause when talking about 7zip source code and "strongly" recommended it be run in a sandbox right when the latest release came out...
ZING
😄
1
Last edit: Hammerfest 2021-03-12
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Does it have support for newer algorithms (e.g. Brotli, Fast-LZMA2, Lizard, LZ4, LZ5 and Zstandard), esp. Zstandard, like https://github.com/mcmilk/7-Zip-zstd
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Igor,
I re-started to recover my 73 GiB archive created earlier with p7zip under Linux now with 7-Zip 21.02 alpha under Win 10. I pushed from the drop down menu 'Open inside #' item and I got 52 different zip files now. (se attached) I can see the offsets and the last zip may contains my zipped stuff.
Can you please provide me with an idea what shall I do to unpack the complete package? Thanks.
Or is there any tricks to concatenate zip file parts and unzip the whole stuff?
don't use this forum thread for that problem.
Read this:
https://www.7-zip.org/recover.html
Hi Igor,
for speed comparison (default setting):
LZMA: ~ 6 MB / s
LZMA2 -mmt8: ~ 11 MB / s
RAR5 -s: ~ 21 MB / s
The sample data is 483 MB.
I have i7-4790K and 7z can use 8 threads. It's too slow for 8 threads.
But it's faster than p7zip.
OS: Xubuntu
Well, I tried with Raspberry Pi4 @2 GHz for armhf(32bit), same settings as with 64-bit:
https://sourceforge.net/p/sevenzip/discussion/45797/thread/d401ab2966/#b089/96da
but it crashed at crc32:32 4t every time. 1t and 2t worked fine, so it may has something to do with overclocking, so here are results for standard clocks with raspian 2021.1 for pi4 4GB:
Last edit: HITCHER 2021-03-10
FX-8300 @3,6GHz (3,9/4,2GHz turbo):
on Debian 11 (bullseye)
Last edit: HITCHER 2021-03-13
The hardware instruction crc32 is very fast on Cortex-A72. Probably it can be executed with latency=1 on A72.
And Cortex-A57 shows 3 cycles latency for same instruction.
So A72 is 3 times faster there. All another things look similar for performance of A72 and A57, but A72 in Pi4 doesn't support SHA and AES instructions.
That CRC32 test is pretty fast on A72 to show the bandwidth of RAM and cache when the block is large.
You can run it on arm64 so:
7zz b -mm=CRC32:64 -mtime=2Maybe you can use similar CRC32 benchmark also to catch exact lines with crash. You can change the time with
-mtime=2for seconds or with-mtimems=100for ms, if you want faster runs.And what is that crash on 2 GHz?
Is it some error message from 7-Zip program or it's some crash on system level?
Last edit: Igor Pavlov 2021-03-10
The crash was black screen and what appears to be an instant shutdown of SOC.
Strange, because on ARM64 it worked stable so far oc 2GHz, but not for this 32bit code.
My pi has heatsink and fan attached, and a very powerful PSU (not usb-c), so this was not the problem. It seems, the problem was GPU OC setting of 650 MHz, i changed that to 550, maybe that has caused a deadlock before (GPU doesn't do anything for 7zz, but on pi it also has the memory controller). Now it's working on 2GHz cpu clock.
On Ubuntu20.10 for Raspberrypi ARM64, you can add architecture armhf to dpkg and to repositories, and install the required libraries, libc6 and libstdc++ to get the 32bit armhf binary working.
https://wiki.debian.org/Multiarch/HOWTO
I have added below raspberrypi4 armhf and the ARM64 benchmarks from the other thread for comparison.
Last edit: HITCHER 2021-03-12
I installed multiarch armhf + ARM64 on ubuntu 20.10 ARM64 for pi4, and was able to run 32 Bit code there too, now even with overclock settings (which crashed on raspbian 32Bit):
Still i think GPU clock 650 is considered to be unstable, since it crashed before on raspian.
Something else is different. On raspian there was 32Bit kernel.
Last edit: HITCHER 2021-03-15
OS: Slackware64 -current
I have been using 7-Zip for 64-bit Linux x86-64 (AMD64) since the first day of this version's release. I replaced the jinfeihan57/p7zip fork in the Amavis configuration on my test server without any problems, nothing recorded in the logs.
+1 for adding Zstandard.
However, I found something in /var/log/syslog
Should I worry?
Last edit: teoberi 2021-04-09
So what's the difference between 7zip and p7zip?
And what are requirements for run except
GLIBC_2.16andCXXABI_1.3.9?I want to know this too, what is the difference? Of course in summary.
Don't know if this is intentional, but the -o switch works in a funny way. Let's say I downloaded the source code for 7z and wanted to extract it in a directory same as the filename of the archive. Adding a space between -o and the output directory breaks the switch.
I'm also uploading my benchmark.
It always worked like that.
Maybe it should be stated in documentation that this option does not allow space after switch.
It's same rule for all switches in 7-Zip.
And you can move switches to any place in command line among non switches.
Just wanted to drop another note of thanks for the native Linux support. Very welcome and surprising news, especially considering p7zip seems to be abandoned at this point.
Last edit: Jared B. 2021-03-11
Does it involve some significant changes in code or just prepared scripts and makefiles necessary to compilation?
I also compiled 7-Zip code with p7zip's scripts for test purposes. So it will be possible to merge both code versions.
Last edit: Igor Pavlov 2021-03-11
Linux 5.11.5-zen1-1-zen x86_64
I've just checked it on Fedora 33 and it worked fine. However, I've noticed that with
-pit prints a given password while typing. Is there a switch to hide it?In addition to that, for
-pthere was a nice feature in p7zip to ask twice for manually typed (and hidden) password. Thanks to that you didn't need to write the password once and then test the archive to verify that you entered the right password. Typing it twice was very effective in my case to eliminate typos. IT would be nice to have it also available in the new version.Arch Linux, 5.11.5-zen1-1-zen x86_64:
https://twitter.com/AdmVonSchneider/status/1369300173441089538
I hope he has reached out to you instead of just posting and claiming to not want to post it publicly.Considering you don't normally release source until its final is this scare mongering (if he hasn't reached out)?
He has since gone on to "elaborate" on his "strongly" worded first post, naturally he didn't understand the power of words with regards to the OBVIOUS worry it would cause when talking about 7zip source code and "strongly" recommended it be run in a sandbox right when the latest release came out...
ZING
Last edit: Hammerfest 2021-03-12
Does it have support for newer algorithms (e.g. Brotli, Fast-LZMA2, Lizard, LZ4, LZ5 and Zstandard), esp. Zstandard, like https://github.com/mcmilk/7-Zip-zstd