Re: Segfault when building openjdk13 with openjdk12
On Thu, 2022-02-03 at 12:31 -0500, Abigail G wrote: > On Wed, 2022-02-02 at 12:12 +0300, Aleksey Shipilev wrote: > > On 2/2/22 08:53, Abigail G wrote: > > > Whoops, looks like I made the zip wrong, this one should work: > > > https://0x0.st/oHxy.zip > > > > So it looks like a GC crash: > > > > # SIGSEGV (0xb) at pc=0x7fa2ba719208, pid=29539, tid=29557 > > > > siginfo: si_signo: 11 (SIGSEGV), si_code: 1 (SEGV_MAPERR), si_addr: > > 0x0100 > > > > Current thread (0x7fa2b4091800): GCTaskThread "GC Thread#0" > > [stack: > > 0x7fa2b8c1,0x7fa2b8d1] [id=29557] > > > > > > If I was stuck with issue like this, I'll try the following: > > > > *) Test the memory. Memory problems frequently manifest as GC > > bugs: > > https://shipilev.net/jvm/test-your-memory/ > > Can confirm it's not memory, I was able to reproduce it on another > machine. > > > > > *) Use fastdebug JDK as the boot JDK, looking for a reasonable > > assert instead of anonymous crash. > > This probably requires building it first after configuring using -- > > with-debug-level=fastdebug. You > > said the build eventually succeeds, so a few tries might be in > > order. > > Will try this. > > > > > *) Bootstrap JDK 13 build with JDK 13. AFAICS, the actual > > requirement for boot JDK is 12..13. > > The problem with this is jdk13 doesn't exist yet in void, so we have > to > do a bootstrap chain up from 11 (the latest version packaged). > > > Regardless, even if you find a bug in JDK 12, it is unlikely to get > > fixed, as no one maintains JDK > > 12. Azul maintains JDK 13, so depending on your luck of identifying > > the issue with JDK 13, I'd say > > fixing JDK 13 might be possible. > > > > Thanks for the tips! > > Abigail G Ok, I build jdk12 with fastdebug and built jdk13 with that, still segfaults but maybe the core dump will have more info now: https://0x0.st/oHkf.zip Abigail G
Re: Segfault when building openjdk13 with openjdk12
On Wed, 2022-02-02 at 12:12 +0300, Aleksey Shipilev wrote: > On 2/2/22 08:53, Abigail G wrote: > > Whoops, looks like I made the zip wrong, this one should work: > > https://0x0.st/oHxy.zip > > So it looks like a GC crash: > > # SIGSEGV (0xb) at pc=0x7fa2ba719208, pid=29539, tid=29557 > > siginfo: si_signo: 11 (SIGSEGV), si_code: 1 (SEGV_MAPERR), si_addr: > 0x0100 > > Current thread (0x7fa2b4091800): GCTaskThread "GC Thread#0" > [stack: > 0x7fa2b8c1,0x7fa2b8d1] [id=29557] > > > If I was stuck with issue like this, I'll try the following: > > *) Test the memory. Memory problems frequently manifest as GC bugs: > https://shipilev.net/jvm/test-your-memory/ Can confirm it's not memory, I was able to reproduce it on another machine. > > *) Use fastdebug JDK as the boot JDK, looking for a reasonable > assert instead of anonymous crash. > This probably requires building it first after configuring using -- > with-debug-level=fastdebug. You > said the build eventually succeeds, so a few tries might be in order. Will try this. > > *) Bootstrap JDK 13 build with JDK 13. AFAICS, the actual > requirement for boot JDK is 12..13. The problem with this is jdk13 doesn't exist yet in void, so we have to do a bootstrap chain up from 11 (the latest version packaged). > Regardless, even if you find a bug in JDK 12, it is unlikely to get > fixed, as no one maintains JDK > 12. Azul maintains JDK 13, so depending on your luck of identifying > the issue with JDK 13, I'd say > fixing JDK 13 might be possible. > Thanks for the tips! Abigail G
Re: Segfault when building openjdk13 with openjdk12
On 2/2/22 08:53, Abigail G wrote: Whoops, looks like I made the zip wrong, this one should work: https://0x0.st/oHxy.zip So it looks like a GC crash: # SIGSEGV (0xb) at pc=0x7fa2ba719208, pid=29539, tid=29557 siginfo: si_signo: 11 (SIGSEGV), si_code: 1 (SEGV_MAPERR), si_addr: 0x0100 Current thread (0x7fa2b4091800): GCTaskThread "GC Thread#0" [stack: 0x7fa2b8c1,0x7fa2b8d1] [id=29557] If I was stuck with issue like this, I'll try the following: *) Test the memory. Memory problems frequently manifest as GC bugs: https://shipilev.net/jvm/test-your-memory/ *) Use fastdebug JDK as the boot JDK, looking for a reasonable assert instead of anonymous crash. This probably requires building it first after configuring using --with-debug-level=fastdebug. You said the build eventually succeeds, so a few tries might be in order. *) Bootstrap JDK 13 build with JDK 13. AFAICS, the actual requirement for boot JDK is 12..13. Regardless, even if you find a bug in JDK 12, it is unlikely to get fixed, as no one maintains JDK 12. Azul maintains JDK 13, so depending on your luck of identifying the issue with JDK 13, I'd say fixing JDK 13 might be possible. -- Thanks, -Aleksey
Re: Segfault when building openjdk13 with openjdk12
On 2/02/2022 3:53 pm, Abigail G wrote: On Wed, 2022-02-02 at 15:48 +1000, David Holmes wrote: Hi Abigail, On 2/02/2022 3:01 pm, Abigail G wrote: Hello, I'm working on packaging several openjdk versions for Void Linux. When I build jdk13 with jdk12 as the boot jdk, I get a segfault very soon into the build. I am not a java developer, so I'm not quite sure what to make of the log, trace, and core dump. Anyone have any ideas? Could just be a bug in 12, fixed later. Interestingly, when I repeatedly restart the build without removing any of the partially-built state, it eventually succeeds (usually after 1-3 retries), and I have been able to use the result to build jdk14 without issue. I've attached a .zip with the relevant build logs, hotspot error log, and core dump (if that doesn't work: https://0x0.st/oHxe.zip). The mailing list strips attachments and the link to the zip file gave a file with a single empty directory in it - no logs. Cheers, David The scripts I'm using to build jdk12 and 13 are available here: https://github.com/classabbyamp/void-packages/blob/openjdk17/srcpkgs/openjdk12-bootstrap/template https://github.com/classabbyamp/void-packages/blob/openjdk17/srcpkgs/openjdk13-bootstrap/template TIA, Abigail G Whoops, looks like I made the zip wrong, this one should work: https://0x0.st/oHxy.zip Unfortunately no symbolic info in the hs-err file so nothing to go on. All I can suggest is that you build the latest 12u to use to bootstrap the latest 13u etc. Cheers, David Abigail G
Re: Segfault when building openjdk13 with openjdk12
On Wed, 2022-02-02 at 15:48 +1000, David Holmes wrote: > Hi Abigail, > > On 2/02/2022 3:01 pm, Abigail G wrote: > > Hello, > > > > I'm working on packaging several openjdk versions for Void Linux. > > > > When I build jdk13 with jdk12 as the boot jdk, I get a segfault > > very > > soon into the build. I am not a java developer, so I'm not quite > > sure > > what to make of the log, trace, and core dump. Anyone have any > > ideas? > > Could just be a bug in 12, fixed later. > > > Interestingly, when I repeatedly restart the build without removing > > any > > of the partially-built state, it eventually succeeds (usually after > > 1-3 > > retries), and I have been able to use the result to build jdk14 > > without > > issue. > > > > I've attached a .zip with the relevant build logs, hotspot error > > log, > > and core dump (if that doesn't work: https://0x0.st/oHxe.zip). > > The mailing list strips attachments and the link to the zip file gave > a > file with a single empty directory in it - no logs. > > Cheers, > David > > > The scripts I'm using to build jdk12 and 13 are available here: > > https://github.com/classabbyamp/void-packages/blob/openjdk17/srcpkgs/openjdk12-bootstrap/template > > https://github.com/classabbyamp/void-packages/blob/openjdk17/srcpkgs/openjdk13-bootstrap/template > > > > TIA, > > > > Abigail G Whoops, looks like I made the zip wrong, this one should work: https://0x0.st/oHxy.zip Abigail G
Re: Segfault when building openjdk13 with openjdk12
Hi Abigail, On 2/02/2022 3:01 pm, Abigail G wrote: Hello, I'm working on packaging several openjdk versions for Void Linux. When I build jdk13 with jdk12 as the boot jdk, I get a segfault very soon into the build. I am not a java developer, so I'm not quite sure what to make of the log, trace, and core dump. Anyone have any ideas? Could just be a bug in 12, fixed later. Interestingly, when I repeatedly restart the build without removing any of the partially-built state, it eventually succeeds (usually after 1-3 retries), and I have been able to use the result to build jdk14 without issue. I've attached a .zip with the relevant build logs, hotspot error log, and core dump (if that doesn't work: https://0x0.st/oHxe.zip). The mailing list strips attachments and the link to the zip file gave a file with a single empty directory in it - no logs. Cheers, David The scripts I'm using to build jdk12 and 13 are available here: https://github.com/classabbyamp/void-packages/blob/openjdk17/srcpkgs/openjdk12-bootstrap/template https://github.com/classabbyamp/void-packages/blob/openjdk17/srcpkgs/openjdk13-bootstrap/template TIA, Abigail G
