Sign in Join

Live chat for Free Pascal, Lazarus, Delphi and every Pascal programmer.

Mailing lists / lazarus / 2023-01

make all compilation error after 319649fb (Attn: Martin)

11 messages · 5 people · –

Reply by email Subscribe to this list Your subscription

Read-only archive. Use j and k to move between messages.

Marcus Sackrow

Hi,

make all for at least x86_64-linux does not compile since 319649fb

https://gitlab.com/freepascal.org/lazarus/lazarus/-/commit/319649fbe62f9eb6798a7d3ffaf976f9f1d31082

the error message:

(3104) Compiling ./frames/initialsetupdlgdebuggerframe.pas /var/lib/jenkins/workspace/LCL_off_x86_64-linux/ide/./frames/initialsetupdlgdebuggerframe.pas(11,3) Fatal: (10022) Can't find unit FpLldbDebugger used by InitialSetupDlgDebuggerFrame Fatal: (1018) Compilation aborted
make[2]: *** [Makefile:4958: lazarus] Error 1

Full log:

https://build.alb42.de:8081/job/LCL_off_x86_64-linux/12240/consoleFull

Could please look someone into it?

Greetings,

Marcus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus-ide.org/pipermail/lazarus/attachments/20230108/d767aee4/attachment.htm>

Reply by email

Wayne Sherman
to Marcus Sackrow
> make all for at least x86_64-linux does not compile since 319649fb thanks it compiles again

About a week ago Lazarus was failing to build with "make useride". Are these two build targets being run as part of the CI / CD tests?

Reply by email

Martin Frb
to Wayne Sherman
Wayne Sherman wrote
>> make all for at least x86_64-linux does not compile since 319649fb > thanks it compiles again About a week ago Lazarus was failing to build with "make useride". Are these two build targets being run as part of the CI / CD tests?

I don't know, it will need to be checked...
The CI config is part of the git repo. So it can be checked by anyone.

Not sure though if running all make targets is the best idea. IMHO more important to run diff OS/Arch targets and compiler versions.

Reply by email

Marcus Sackrow
to Wayne Sherman: About a week ago Lazarus was failing to build with "make useride". Are these two build…

Hi,

Wayne Sherman wrote
>> make all for at least x86_64-linux does not compile since 319649fb > thanks it compiles again About a week ago Lazarus was failing to build with "make useride". Are these two build targets being run as part of the CI / CD tests?

these compilations (from this error report) are on m own Jenkins server and atm it does only a "make all"

it's mainly focused on catching problems with latest FPC (because it uses always the latest FPC, compiled some minutes before that)

and to check if the compilation for Amiga-style systems still work (and with it the MUI widget set) in other jobs

Of course if there is interest in more tasks, it can be added, the server still has some resources.

Greetings,

Marcus

Reply by email

Martin Frb
to Marcus Sackrow
Marcus Sackrow wrote 20 lines
Hi, Am 08.01.23 um 23:11 schrieb Wayne Sherman: >>> make all for at least x86_64-linux does not compile since 319649fb >> thanks it compiles again > About a week ago Lazarus was failing to build with "make useride". > Are these two build targets being run as part of the CI / CD tests? these compilations (from this error report) are on m own Jenkins server and atm it does only a "make all" it's mainly focused on catching problems with latest FPC (because it uses always the latest FPC, compiled some minutes before that) and to check if the compilation for Amiga-style systems still work (and with it the MUI widget set) in other jobs Of course if there is interest in more tasks, it can be added, the server still has some resources.

It needs to be discussed with other devs as well, what they think...

We do still have free the amount of combinations
for testing is (near) infinite.

There is "make" (aka "make all" ?) and "make bigide". Afaik "make useride" depends on what packages are configured, so that is not just a single setting?

But each of them may fail only for a specific OS/Arch, or for some other specific setting. (like compile time range check error for constant eval, with -Cr enabled).
We already have builds for win, gtk2, qt, iirc some with 32bit, and fpc 3.2.2 and trunk. (not all variations though).

I personally currently don't have the time to look into it. (I'll ping the others, if anyone wants to pick up)

If I should find time eventually, my interest will be much more in adding some of the tests we have to be run.

Reply by email

Mattias Gaertner
to Martin Frb
[...] There is "make" (aka "make all" ?) and "make bigide". Afaik "make useride" depends on what packages are configured, so that is not just a single setting?

Yes. But if you start with an empty config directory "make useride" should be reliable and is a good test for lazbuild.

[...]

Mattias

Reply by email

Maxim Ganetsky
to Martin Frb: It needs to be discussed with other devs as well, what they think... We do still have…

09.01.2023 20:01, Martin Frb via lazarus ?????:

Show quoted text 26 lines
On 09/01/2023 17:50, Marcus Sackrow via lazarus wrote: > Hi, > > Am 08.01.23 um 23:11 schrieb Wayne Sherman: >>>> make all for at least x86_64-linux does not compile since 319649fb >>> thanks it compiles again >> About a week ago Lazarus was failing to build with "make useride". >> Are these two build targets being run as part of the CI / CD tests? > > these compilations (from this error report) are on m own Jenkins server > and atm it does only a "make all" > > it's mainly focused on catching problems with latest FPC (because it > uses always the latest FPC, compiled some minutes before that) > > and to check if the compilation for Amiga-style systems still work (and > with it the MUI widget set) in other jobs > > > Of course if there is interest in more tasks, it can be added, the > server still has some resources. It needs to be discussed with other devs as well, what they think... We do still have free the amount of combinations for testing is (near) infinite.

Yes.

Show quoted text 9 lines
There is "make" (aka "make all" ?) and "make bigide". Afaik "make useride" depends on what packages are configured, so that is not just a single setting? But each of them may fail only for a specific OS/Arch, or for some other specific setting. (like compile time range check error for constant eval, with -Cr enabled). We already have builds for win, gtk2, qt, iirc some with 32bit, and fpc 3.2.2 and trunk. (not all variations though).

Yes, details are here:
https://wiki.freepascal.org/Lazarus_Team_-_Git_Workflows#Continuous_integration_.28CI.29_System

There are too much variations and I don't think that it is feasible to define them all.

I think that current set of jobs in CI is quite comprehensible, but if there is justified need, yet another job can be added indeed.

Show quoted text 5 lines
I personally currently don't have the time to look into it. (I'll ping the others, if anyone wants to pick up) If I should find time eventually, my interest will be much more in adding some of the tests we have to be run.
Best regards, Maxim Ganetsky mailto:

Reply by email

Maxim Ganetsky
to Mattias Gaertner: On Mon, 9 Jan 2023 18:01:19 +0100 Yes. But if you start with an empty config directory…

09.01.2023 21:14, Mattias Gaertner via lazarus ?????:

Show quoted text 10 lines
On Mon, 9 Jan 2023 18:01:19 +0100 Martin Frb via lazarus wrote: > [...] > There is "make" (aka "make all" ?) and "make bigide". > Afaik "make useride" depends on what packages are configured, so that > is not just a single setting? Yes. But if you start with an empty config directory "make useride" should be reliable and is a good test for lazbuild.

It is no problem to add some `make useride` job. In principle, we can even switch e. g. x86_64-win64 job to this (to avoid crowding even more our already huge job list). What do you think?

Best regards, Maxim Ganetsky mailto:

Reply by email

Wayne Sherman
to Maxim Ganetsky
make all for at least x86_64-linux does not compile since 319649fb
There is "make" (aka "make all" ?) and "make bigide".

If there is an automatic job for "make all", it apparently did not catch the build failure reported by Marcus.

PM Maxim Ganetsky wrote
It is no problem to add some `make useride` job. In principle, we can even switch e. g. x86_64-win64 job to this (to avoid crowding even more our already huge job list).

Another option, if you don't want to create more jobs, is to string build tests together on the same job:
#pseudocode
make all || echo 'make all failed'; exit 1
make distclean || echo 'make distclean failed'; exit 2 make useride || echo 'make useride failed'; exit 3

Reply by email

Maxim Ganetsky
to Wayne Sherman

10.01.2023 3:37, Wayne Sherman ?????:

Show quoted text 8 lines
On Sun, Jan 8, 2023 Marcus Sackrow wrote: > make all for at least x86_64-linux does not compile since 319649fb On Mon, Jan 9, 2023 Martin Frb wrote: > There is "make" (aka "make all" ?) and "make bigide". If there is an automatic job for "make all", it apparently did not catch the build failure reported by Marcus.

We have only `make bigide` ones, otherwise it would have failed indeed.

Best regards, Maxim Ganetsky mailto:

Reply by email