web/import.inc.php: account for failures that contain an images/ path
PATCH 1/1 web/import.inc.php: account for failures that contain an images/ path
With the testing of BR2_REPRODUCIBLE=y recently added and its generation of a tar filesystem, we are now testing the tar filesystem generation logic, and this reveals some interesting breakage, such as:
http://autobuild.buildroot.net/results/fe3/fe378bca29c86a681ba9ad40386cb89248195c50/build-end.log
However, the "reason" deduced from the build log by the PHP logic is a bit crappy, it's "/home/br-user/autobuild/run/instance-0/output/images/rootfs.tar".
This commit adjusts the ugly PHP code with even more ugliness to take into account the fact that a path may look like foo/thing/build/<pkg>-<version>/.stamp_something or like foo/thing/images/rootfs.tar.
The reasons extracted for the two previous examples would be "<pkg>-<version>" (like is done today) and "rootfs.tar".
--- web/import.inc.php | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
web/import.inc.php +1 -1
@@ -231,7 +231,7 @@ function import_result($buildid, $filename) $reason = "none"; else { $tmp = Array(); - exec("tail -3 " . $thisbuildfinaldir . "build-end.log | grep -v '\[_all\]' | grep 'make.*: \*\*\*' | sed 's,.*\[\([^\]*\)\] Error.*,\\1,' | sed 's,.*/build/\([^/]*\)/.*,\\1,'", $tmp); + exec("tail -3 " . $thisbuildfinaldir . "build-end.log | grep -v '\[_all\]' | grep 'make.*: \*\*\*' | sed 's,.*\[\([^\]*\)\] Error.*,\\1,' | sed 's,.*/\(images\|build\)/\([^/]*\).*,\\2,'", $tmp); if (trim($tmp[0])) $reason = $tmp[0]; else {
Signature · 1 line
2.21.0
PATCH 1/1 web/import.inc.php: account for failures that contain an images/ path
Quoted 11 lines · Thomas Petazzoni
With the testing of BR2_REPRODUCIBLE=y recently added and its generation of a tar filesystem, we are now testing the tar filesystem generation logic, and this reveals some interesting breakage, such as:
http://autobuild.buildroot.net/results/fe3/fe378bca29c86a681ba9ad40386cb89248195c50/build-end.log
However, the "reason" deduced from the build log by the PHP logic is a bit crappy, it's "
/home/br-user/autobuild/run/instance-0/output/images/rootfs.tar".This commit adjusts the ugly PHP code with even more ugliness to take
This sounds like a great idea :-)
Note that the Python code already has similar logic to extract the reason. Why
don't we save that logic as part of the tarball in a "reason" file, and extract it from there?
I know someone who has been working on the autobuild_run script lately who
could certainly do that :-)
That said, we still have to support the legacy database which doesn't have that
"reason" file. So it's still useful to apply this patch.
Regards,
Arnout
Quoted 26 lines
into account the fact that a path may look like foo/thing/build/<pkg>-<version>/.stamp_something or like foo/thing/images/rootfs.tar.
The reasons extracted for the two previous examples would be "<pkg>-<version>" (like is done today) and "rootfs.tar".
Signed-off-by Thomas Petazzoni <[email protected]>--- web/import.inc.php | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)web/import.inc.php +1
-1@@ -231,7 +231,7 @@ function import_result($buildid, $filename) $reason = "none"; else { $tmp = Array(); - exec("tail -3 " . $thisbuildfinaldir . "build-end.log | grep -v '\[_all\]' | grep 'make.*: \*\*\*' | sed 's,.*\[\([^\]*\)\] Error.*,\\1,' | sed 's,.*/build/\([^/]*\)/.*,\\1,'", $tmp); + exec("tail -3 " . $thisbuildfinaldir . "build-end.log | grep -v '\[_all\]' | grep 'make.*: \*\*\*' | sed 's,.*\[\([^\]*\)\] Error.*,\\1,' | sed 's,.*/\(images\|build\)/\([^/]*\).*,\\2,'", $tmp); if (trim($tmp[0])) $reason = $tmp[0]; else {
PATCH 1/1 web/import.inc.php: account for failures that contain an images/ path
Hello,
Quoted 3 lines · On Mon, 17 Jun 2019 21:39:22 +0200 Arnout Vandecappelle <[email protected]> wrote:
Quoted 1 line
This commit adjusts the ugly PHP code with even more ugliness to take
This sounds like a great idea :-)
Isn't it :-)
Quoted 3 lines
Note that the Python code already has similar logic to extract the reason. Why
don't we save that logic as part of the tarball in a "reason" file, and extract it from there?
Because history, but indeed it would make a lot more sense to extract the reason only on the client side, because we anyway extract it already to be able to extract the relevant part of the build log.
Quoted 2 lines
That said, we still have to support the legacy database which doesn't have that
"reason" file. So it's still useful to apply this patch.
There is really no "legacy" to support here. On the server side, when a result is submitted, the reason is extracted from the build log, and then stored in the SQL database. Once that is done, when browsing build results, the reason is always taken from the SQL database.
So server-side, we can entirely drop the logic that calculates the reason from the build log, and instead replace it with just reading the reason from an additional "reason" file in the result tarball submitted by build slaves.
What needs to be done however is a transition period:
- Adding the "reason" file to the build results on the client side is implemented.
- This change is deployed to all build slaves.
- Once it's deployed to all build slaves, we can change the server side to use the reason file instead of calculating the reason manually.
Or we could avoid that transition by using the reason file if available, and fall back to the old logic if not available.
In the mean time, should I still apply this patch ? :-)
Thanks,
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Bootlin Embedded Linux and Kernel engineering https://bootlin.com
PATCH 1/1 web/import.inc.php: account for failures that contain an images/ path
Quoted 19 lines · Thomas Petazzoni
Hello,
Quoted 3 lines · On Mon, 17 Jun 2019 21:39:22 +0200 Arnout Vandecappelle <[email protected]> wrote:
Quoted 1 line
This commit adjusts the ugly PHP code with even more ugliness to take
This sounds like a great idea :-)
Isn't it :-)
Quoted 3 lines
Note that the Python code already has similar logic to extract the reason. Why
don't we save that logic as part of the tarball in a "reason" file, and extract it from there?Because history, but indeed it would make a lot more sense to extract the reason only on the client side, because we anyway extract it already to be able to extract the relevant part of the build log.
Quoted 7 lines
Quoted 2 lines
That said, we still have to support the legacy database which doesn't have that
"reason" file. So it's still useful to apply this patch.There is really no "legacy" to support here. On the server side, when a result is submitted, the reason is extracted from the build log, and then stored in the SQL database. Once that is done, when browsing build results, the reason is always taken from the SQL database.
OK, I thought this code was executed on every query. But that would indeed be
crazy.
Quoted 10 lines
So server-side, we can entirely drop the logic that calculates the reason from the build log, and instead replace it with just reading the reason from an additional "reason" file in the result tarball submitted by build slaves.
What needs to be done however is a transition period:
- Adding the "reason" file to the build results on the client side is implemented.
Quoted 8 lines
- This change is deployed to all build slaves.
- Once it's deployed to all build slaves, we can change the server side to use the reason file instead of calculating the reason manually.
Or we could avoid that transition by using the reason file if available, and fall back to the old logic if not available.
This was the path I was assumed was going to be taken.
Quoted 1 line
In the mean time, should I still apply this patch ? :-)
Depends a bit on how long it's going to take Atharva to implement this reason
file. Atharva?
Regards,
Arnout
PATCH 1/1 web/import.inc.php: account for failures that contain an images/ path
Quoted 1 line · Arnout Vandecappelle
Depends a bit on how long it's going to take Atharva to implement this
reason
Quoted 1 line
file. Atharva?
I will complete the transition to builder-class by end of tomorrow. Then I'll work on the
reason file implementation. Should be done in a day or two I think. So should be ready by
the end of 21st June. Does that sound like a reasonable timeframe?
Regards,
Atharva Lele
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20190618/177678d4/attachment.html>