first-setup.sh from ConTeXt standalone contains the following lines: readelf -A /proc/self/exe | grep -q '^ \+Tag_ABI_VFP_args' if [ ! $? ]; then That probably does not work as intended, because [ ! $? ] is equivalent to [ ! -n $? ]. It does not return the negation of the return code of the previous command. [ ! 0 ] and [ ! 1 ] both have exit code 1, because the strings '0' and '1' are both non-null. Probably no-one actually uses ARMv7 with softfp, including me.
On Mon, Nov 5, 2018 at 6:14 PM Sebastian Miele
first-setup.sh from ConTeXt standalone contains the following lines:
readelf -A /proc/self/exe | grep -q '^ \+Tag_ABI_VFP_args' if [ ! $? ]; then
That probably does not work as intended, because [ ! $? ] is equivalent to [ ! -n $? ]. It does not return the negation of the return code of the previous command. [ ! 0 ] and [ ! 1 ] both have exit code 1, because the strings '0' and '1' are both non-null.
Probably no-one actually uses ARMv7 with softfp, including me.
Ok, thank you very much. (I think that the script could be something like readelf -A /proc/self/exe | grep -q '^ \+Tag_ABI_VFP_args' && true || platform="linux-armel" ) Are you using 32bit linux-gnueabi ? -- luigi
On Mon, Nov 5, 2018 at 9:44 PM luigi scarso
[..]
Are you using 32bit linux-gnueabi ?
I am tinkering with an Alpine Linux chroot on my Samsung Galaxy Note 2 which has a custom rom flashed. Alpine Linux uses musl libc. First I wanted to go with a distro based on glibc (Arch Linux ARM), but that did not work, because recent versions of glibc require a Kernel more recent that what is available for the Note 2. Inside the Alpine Linux chroot 'uname -a' says: Linux auge 3.0.101-ReTRoKeRNeL-g6242745fe35 #3 SMP PREEMPT Fri Aug 3 16:31:55 EEST 2018 armv7l Linux. first-setup.sh distinguishes between musl libc and glibc only for Linux on i*86 and x86_64|ia64, and not for Linux on arm*. After rsync -rlptv rsync://contextgarden.net/minimals/setup/linux-armhf/bin . the command 'ls -l "$PWD/bin/luatex"' returns: -rwxr-xr-x 1 w w 5720504 Oct 25 08:16 /home/w/borg/tex/con/bin/luatex which is correct. But "$PWD/bin/luatex --version" returns the rather strange error: zsh: no such file or directory: /home/w/borg/tex/con/bin/luatex It probably has to do with the circumstance that 'readelf -a "$PWD/bin/luatex" | grep -i glibc' succeeds, while 'readelf -a /bin/busybox | grep -i glibc' does not. I probably do not use linux-gnueabi, but something like armv7a-unknown-linux-musleabihf, although I still find this stuff very confusing. This was just an experiment. I do not really need this.
Dear Sebastian, On Mon, 5 Nov 2018 at 18:14, Sebastian Miele wrote:
first-setup.sh from ConTeXt standalone contains the following lines:
readelf -A /proc/self/exe | grep -q '^ \+Tag_ABI_VFP_args' if [ ! $? ]; then
That probably does not work as intended, because [ ! $? ] is equivalent to [ ! -n $? ]. It does not return the negation of the return code of the previous command. [ ! 0 ] and [ ! 1 ] both have exit code 1, because the strings '0' and '1' are both non-null.
I'm sorry. A few years back we had a long discussion with the consensus that we should have used if [ $? != 0 ]; then instead. I have no idea why this wasn't the case at the end. Maybe I just wasn't reading careful enough?
Probably no-one actually uses ARMv7 with softfp, including me.
We kept building the armel binaries on some super strange hardware configuration which Boris provided to us. At some point he was no longer able to provide the hardware, we no longer built the binaries & provided them, and I don't think that anybody complained since. The armhf binaries are built on RPi in Hans' cellar running Raspbian. There's a high probability that some similar code is used in TeX Live as well, but I would need to check. (But maybe Karl fixed it correctly, only I screwed up :) Mojca
On Tue, Nov 6, 2018 at 7:45 AM Mojca Miklavec
A few years back we had a long discussion with the consensus that we should have used
if [ $? != 0 ]; then
This seems to have been fixed in the mean time.
We kept building the armel binaries on some super strange hardware configuration which Boris provided to us. At some point he was no longer able to provide the hardware, we no longer built the binaries & provided them, and I don't think that anybody complained since.
The armhf binaries are built on RPi in Hans' cellar running Raspbian.
As noted in the previous mail I do not really need it in the foreseeable future. Apart from that there are additional problems like glibc vs. musl libc and maybe different versions of floating point units. If I or someone else really needs it in the future I would rather try to adopt what is running on the RPi in Hans' cellar to variable qemu guest and/or to variable cross compilation toolchains.
On Thu, 8 Nov 2018 at 22:21, Sebastian Miele wrote:
On Tue, Nov 6, 2018 at 7:45 AM Mojca Miklavec wrote:
A few years back we had a long discussion with the consensus that we should have used
if [ $? != 0 ]; then
This seems to have been fixed in the mean time.
Yes, I fixed it after you notified us about the problem (thank you).
We kept building the armel binaries on some super strange hardware configuration which Boris provided to us. At some point he was no longer able to provide the hardware, we no longer built the binaries & provided them, and I don't think that anybody complained since.
The armhf binaries are built on RPi in Hans' cellar running Raspbian.
As noted in the previous mail I do not really need it in the foreseeable future. Apart from that there are additional problems like glibc vs. musl libc and maybe different versions of floating point units.
If I or someone else really needs it in the future I would rather try to adopt what is running on the RPi in Hans' cellar to variable qemu guest and/or to variable cross compilation toolchains.
We could run a cross-compiler if needed. There are too many different platforms to cover them all, but if there is a reasonable demand (note that we do have a very small number of users for some platforms), we can try to figure out if we could add it. Regarding musl: support for musl binaries has only been added recently. I did not attempt doing the same check on arm for various reasons, the most important one being that we don't even have the binaries for it. (I could add a check and prevent the default binaries to work, but we could not support the platform out of the box anyway until we add everything that's needed.) Let me know if you think there is still something we should do now. Mojca
participants (3)
-
luigi scarso -
Mojca Miklavec -
Sebastian Miele