iotests: Work around failing readlink -f
PATCH 1/1 iotests: Work around failing readlink -f
On macOS, (out of the box) readlink does not have -f. If the recent "readlink -f" call introduced by b1cbc33a397 fails, just fall back to the old behavior (which means you can run the iotests only from the build tree, but that worked fine for six years, so it should be fine still).
Keep any potential error message on stderr. If users want to run the iotests from outside the build tree, this may point them to what's wrong (with their system).
("iotests: Allow running from different directory")
---
Hi Thomas,
I thought this would be quicker than writing a witty response on whether you or me should write this patch. O:)
--- tests/qemu-iotests/check | 4 ++++ 1 file changed, 4 insertions(+)
tests/qemu-iotests/check +4 -0
@@ -45,6 +45,10 @@ then fi source_iotests=$(cd "$source_iotests"; pwd) || _init_error "failed to enter source tree" build_iotests=$(readlink -f $(dirname "$0")) + if [ "$?" -ne 0 ]; then + # Perhaps -f is unsupported, revert to pre-b1cbc33a397 behavior + build_iotests=$PWD + fi else # called from the source tree source_iotests=$PWD
Signature · 1 line
2.26.2
Quoted 10 lines · Max Reitz
On macOS, (out of the box) readlink does not have -f. If the recent "readlink -f" call introduced by
b1cbc33a397fails, just fall back to the old behavior (which means you can run the iotests only from the build tree, but that worked fine for six years, so it should be fine still).Keep any potential error message on stderr. If users want to run the iotests from outside the build tree, this may point them to what's wrong (with their system).
Quoted 5 lines
Fixes b1cbc33a3971b6bb005d5ac3569feae35a71de0f("iotests: Allow running from different directory")
Reported-by Claudio Fontana <[email protected]>Reported-by Thomas Huth <[email protected]>Signed-off-by Max Reitz <[email protected]>
Queued to testing/next, thanks.
Signature · 1 line
Alex Bennée
Quoted 29 lines · Max Reitz
On macOS, (out of the box) readlink does not have -f. If the recent "readlink -f" call introduced by
b1cbc33a397fails, just fall back to the old behavior (which means you can run the iotests only from the build tree, but that worked fine for six years, so it should be fine still).Keep any potential error message on stderr. If users want to run the iotests from outside the build tree, this may point them to what's wrong (with their system).
Fixes b1cbc33a3971b6bb005d5ac3569feae35a71de0f("iotests: Allow running from different directory")
Reported-by Claudio Fontana <[email protected]>Reported-by Thomas Huth <[email protected]>Signed-off-by Max Reitz <[email protected]>---
Hi Thomas,I thought this would be quicker than writing a witty response on whether you or me should write this patch. O:)
--- tests/qemu-iotests/check | 4 ++++ 1 file changed, 4 insertions(+)tests/qemu-iotests/check +0
-0
tests/qemu-iotests/check then · line 54
Quoted 10 lines
tests/qemu-iotests/check +4
-0@@ -45,6 +45,10 @@ then fi source_iotests=$(cd "$source_iotests"; pwd) || _init_error "failed to enter source tree" build_iotests=$(readlink -f $(dirname "$0")) + if [ "$?" -ne 0 ]; then + # Perhaps -f is unsupported, revert to pre-b1cbc33a397 behavior + build_iotests=$PWD + fi else # called from the source tree source_iotests=$PWD
Thanks, the macOS build seems to work again:
https://cirrus-ci.com/task/5431828267925504?command=main#L7033
Quoted 30 lines · Max Reitz
On macOS, (out of the box) readlink does not have -f. If the recent "readlink -f" call introduced by
b1cbc33a397fails, just fall back to the old behavior (which means you can run the iotests only from the build tree, but that worked fine for six years, so it should be fine still).Keep any potential error message on stderr. If users want to run the iotests from outside the build tree, this may point them to what's wrong (with their system).
Fixes b1cbc33a3971b6bb005d5ac3569feae35a71de0f("iotests: Allow running from different directory")
Reported-by Claudio Fontana <[email protected]>Reported-by Thomas Huth <[email protected]>Signed-off-by Max Reitz <[email protected]>---
Hi Thomas,I thought this would be quicker than writing a witty response on whether you or me should write this patch. O:)
--- tests/qemu-iotests/check | 4 ++++ 1 file changed, 4 insertions(+)tests/qemu-iotests/check +0
-0
tests/qemu-iotests/check then · line 52
Quoted 8 lines
tests/qemu-iotests/check +4
-0@@ -45,6 +45,10 @@ then fi source_iotests=$(cd "$source_iotests"; pwd) || _init_error "failed to enter source tree" build_iotests=$(readlink -f $(dirname "$0")) + if [ "$?" -ne 0 ]; then + # Perhaps -f is unsupported, revert to pre-b1cbc33a397 behavior + build_iotests=$PWD + fi else
This still prints
readlink: illegal option -- f
usage: readlink [-n] [file ...]
(you can see it in the build log that Thomas links to).
build_iotests=$(readlink -f $(dirname "$0") 2>/dev/null)
should avoid that, I think.
thanks
-- PMM
Quoted 44 lines · Peter Maydell
Quoted 30 lines · Max Reitz
On macOS, (out of the box) readlink does not have -f. If the recent "readlink -f" call introduced by
b1cbc33a397fails, just fall back to the old behavior (which means you can run the iotests only from the build tree, but that worked fine for six years, so it should be fine still).Keep any potential error message on stderr. If users want to run the iotests from outside the build tree, this may point them to what's wrong (with their system).
Fixes b1cbc33a3971b6bb005d5ac3569feae35a71de0f("iotests: Allow running from different directory")
Reported-by Claudio Fontana <[email protected]>Reported-by Thomas Huth <[email protected]>Signed-off-by Max Reitz <[email protected]>---
Hi Thomas,I thought this would be quicker than writing a witty response on whether you or me should write this patch. O:)
--- tests/qemu-iotests/check | 4 ++++ 1 file changed, 4 insertions(+)tests/qemu-iotests/check +0
-0tests/qemu-iotests/check then · line 52
Quoted 8 lines
tests/qemu-iotests/check +4
-0@@ -45,6 +45,10 @@ then fi source_iotests=$(cd "$source_iotests"; pwd) || _init_error "failed to enter source tree" build_iotests=$(readlink -f $(dirname "$0")) + if [ "$?" -ne 0 ]; then + # Perhaps -f is unsupported, revert to pre-b1cbc33a397 behavior + build_iotests=$PWD + fi elseThis still prints
readlink: illegal option -- f
usage: readlink [-n] [file ...]
Quoted 5 lines
(you can see it in the build log that Thomas links to).
build_iotests=$(readlink -f $(dirname "$0") 2>/dev/null)should avoid that, I think.
I mentioned in the commit message that I find this useful and desirable behavior. Something isn’t working that perhaps users are expecting to work (because it will work on other systems), so I don’t think the error message should be suppressed.
Max
Quoted 48 lines · Max Reitz
Quoted 46 lines · Peter Maydell
Quoted 30 lines · Max Reitz
On macOS, (out of the box) readlink does not have -f. If the recent "readlink -f" call introduced by
b1cbc33a397fails, just fall back to the old behavior (which means you can run the iotests only from the build tree, but that worked fine for six years, so it should be fine still).Keep any potential error message on stderr. If users want to run the iotests from outside the build tree, this may point them to what's wrong (with their system).
Fixes b1cbc33a3971b6bb005d5ac3569feae35a71de0f("iotests: Allow running from different directory")
Reported-by Claudio Fontana <[email protected]>Reported-by Thomas Huth <[email protected]>Signed-off-by Max Reitz <[email protected]>---
Hi Thomas,I thought this would be quicker than writing a witty response on whether you or me should write this patch. O:)
--- tests/qemu-iotests/check | 4 ++++ 1 file changed, 4 insertions(+)tests/qemu-iotests/check +0
-0tests/qemu-iotests/check then · line 52
Quoted 8 lines
tests/qemu-iotests/check +4
-0@@ -45,6 +45,10 @@ then fi source_iotests=$(cd "$source_iotests"; pwd) || _init_error "failed to enter source tree" build_iotests=$(readlink -f $(dirname "$0")) + if [ "$?" -ne 0 ]; then + # Perhaps -f is unsupported, revert to pre-b1cbc33a397 behavior + build_iotests=$PWD + fi elseThis still prints
readlink: illegal option -- f
usage: readlink [-n] [file ...](you can see it in the build log that Thomas links to).
Quoted 8 lines
Quoted 3 lines
build_iotests=$(readlink -f $(dirname "$0") 2>/dev/null)should avoid that, I think.
I mentioned in the commit message that I find this useful and desirable behavior. Something isn’t working that perhaps users are expecting to work (because it will work on other systems), so I don’t think the error message should be suppressed.
I disagree. Either iotests can handle readlink not having '-f', in which case it should not let readlink spew junk to the log, or it can't, in which case it should bail out.
If you want to tell the user something, you should catch the failure and print something yourself, because then you can do so with a message that will make sense to somebody trying to run the test suite and point them in the direction of what they can do to deal with the problem, eg something like
"readlink version doesn't support '-f'. Assuming that iotests
are being run from the build tree. If this isn't true then we will fail later on: you can either run from the build tree, or install a newer readlink."
(fix up the details to suit, or throw away entirely in favour of some other text if you like).
Printing "readlink: illegal option" is just going to cause people to assume QEMU's scripts are broken and send us bug reports, so please don't do that.
thanks
-- PMM
Quoted 53 lines · Peter Maydell
Quoted 52 lines · Max Reitz
Quoted 49 lines · Peter Maydell
Quoted 30 lines · Max Reitz
On macOS, (out of the box) readlink does not have -f. If the recent "readlink -f" call introduced by
b1cbc33a397fails, just fall back to the old behavior (which means you can run the iotests only from the build tree, but that worked fine for six years, so it should be fine still).Keep any potential error message on stderr. If users want to run the iotests from outside the build tree, this may point them to what's wrong (with their system).
Fixes b1cbc33a3971b6bb005d5ac3569feae35a71de0f("iotests: Allow running from different directory")
Reported-by Claudio Fontana <[email protected]>Reported-by Thomas Huth <[email protected]>Signed-off-by Max Reitz <[email protected]>---
Hi Thomas,I thought this would be quicker than writing a witty response on whether you or me should write this patch. O:)
--- tests/qemu-iotests/check | 4 ++++ 1 file changed, 4 insertions(+)tests/qemu-iotests/check +0
-0tests/qemu-iotests/check then · line 52
Quoted 8 lines
tests/qemu-iotests/check +4
-0@@ -45,6 +45,10 @@ then fi source_iotests=$(cd "$source_iotests"; pwd) || _init_error "failed to enter source tree" build_iotests=$(readlink -f $(dirname "$0")) + if [ "$?" -ne 0 ]; then + # Perhaps -f is unsupported, revert to pre-b1cbc33a397 behavior + build_iotests=$PWD + fi elseThis still prints
readlink: illegal option -- f
usage: readlink [-n] [file ...](you can see it in the build log that Thomas links to).
build_iotests=$(readlink -f $(dirname "$0") 2>/dev/null)should avoid that, I think.
Quoted 8 lines
Quoted 4 lines
I mentioned in the commit message that I find this useful and desirable behavior. Something isn’t working that perhaps users are expecting to work (because it will work on other systems), so I don’t think the error message should be suppressed.
I disagree. Either iotests can handle readlink not having '-f', in which case it should not let readlink spew junk to the log, or it can't, in which case it should bail out.
I find this a bit complicated, because we can’t exactly handle readlink without -f. We can fall back to the old behavior on such systems, which I think is good enough and I assume you think, too.
Quoted 9 lines
If you want to tell the user something, you should catch the failure and print something yourself, because then you can do so with a message that will make sense to somebody trying to run the test suite and point them in the direction of what they can do to deal with the problem, eg something like
"readlink version doesn't support '-f'. Assuming that iotests
are being run from the build tree. If this isn't true then we will fail later on: you can either run from the build tree, or install a newer readlink."
Doesn’t sound bad, it isn’t “bail out”, though, so I don’t fully understand how this relates to your paragraph above. (Because it seems like you suggest printing this unconditionally. I think it would be better to print it only if readlink -f failed, and then check finds $PWD isn’t $build_tree/tests/qemu-iotests. But...)
Quoted 3 lines
Printing "readlink: illegal option" is just going to cause people to assume QEMU's scripts are broken and send us bug reports, so please don't do that.
(That’s absolutely true.)
...given everything, I think the best is then to indeed just suppress readlink’s error message. If readlink doesn’t work, and build_iotests defaults to $PWD, and $PWD is not the build tree, then you’ll end up with the six-year-old error message “(make sure the qemu-iotests are run from tests/qemu-iotests in the build tree)”. Because doing so is probably easier anyway than trying to find a readlink that works.
Max