Segfault when calling PgfiInitPhase2

[sjhalayka]

I am using PGENLIB.

I have a code to read the data from the input files, but it segfaults when I call PgfiInitPhase2 on line 55 of main.cpp in the repository https://github.com/sjhalayka/pgenlib_reader

I compile using the command: g++ *.cc main.cpp -fpermissive

Are there any glaring errors in main.cpp that I'm missing? Thanks for any time that you can provide. I changed line 56 from 1 to 0, but it doesn't help. I also tried dummy files from plink2, instead of my files, and it still segfaults.

[chrchang]

Second-to-last parameter needs to be a uintptr_t pointer, to accept the pgr_alloc_cacheline_ct return value.

[sjhalayka]

Thank you, that helps a lot.

…

On Tue, Mar 4, 2025 at 8:42 AM Christopher Chang ***@***.***> wrote: Second-to-last parameter needs to be a uintptr_t pointer, to accept the pgr_alloc_cacheline_ct return value. — Reply to this email directly, view it on GitHub <#293 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AD5FASQRTQZ7IP65TDDBA5T2SW3XHAVCNFSM6AAAAABYJV4PNGVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTEMZYHE2DGMQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[sjhalayka]

P.S. The program works great where variant count is less than the chunk size. Otherwise, it does a core dump now. I'll take it from here now though -- I appreciate your time up to here.

…

On Tue, Mar 4, 2025 at 9:06 AM Shawn Halayka ***@***.***> wrote: Thank you, that helps a lot. On Tue, Mar 4, 2025 at 8:42 AM Christopher Chang ***@***.***> wrote: > Second-to-last parameter needs to be a uintptr_t pointer, to accept the > pgr_alloc_cacheline_ct return value. > > — > Reply to this email directly, view it on GitHub > <#293 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AD5FASQRTQZ7IP65TDDBA5T2SW3XHAVCNFSM6AAAAABYJV4PNGVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTEMZYHE2DGMQ> > . > You are receiving this because you authored the thread.Message ID: > ***@***.***> >

[sjhalayka]

It's working now. As far as I can tell, I was not allocating enough space for the genovecs array. https://github.com/sjhalayka/pgenlib_reader

…

On Tue, Mar 4, 2025 at 9:09 AM Shawn Halayka ***@***.***> wrote: P.S. The program works great where variant count is less than the chunk size. Otherwise, it does a core dump now. I'll take it from here now though -- I appreciate your time up to here. On Tue, Mar 4, 2025 at 9:06 AM Shawn Halayka ***@***.***> wrote: > Thank you, that helps a lot. > > On Tue, Mar 4, 2025 at 8:42 AM Christopher Chang < > ***@***.***> wrote: > >> Second-to-last parameter needs to be a uintptr_t pointer, to accept the >> pgr_alloc_cacheline_ct return value. >> >> — >> Reply to this email directly, view it on GitHub >> <#293 (comment)>, >> or unsubscribe >> <https://github.com/notifications/unsubscribe-auth/AD5FASQRTQZ7IP65TDDBA5T2SW3XHAVCNFSM6AAAAABYJV4PNGVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTEMZYHE2DGMQ> >> . >> You are receiving this because you authored the thread.Message ID: >> ***@***.***> >> >

[chrchang]

fyi, the genovecs array should have NypCtToWordCt(sample_ct) elements.

[sjhalayka]

I tried that and it gives me a core dump because genovec is too small. Thanks for helping though! I sincerely appreciate it.

…

On Tue, Mar 4, 2025 at 11:52 PM Christopher Chang ***@***.***> wrote: fyi, the genovecs array should have NypCtToWordCt(sample_ct) elements. — Reply to this email directly, view it on GitHub <#293 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AD5FASU3SSADW57HVSIP7UL2S2GJNAVCNFSM6AAAAABYJV4PNGVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTEMZZGY4DEMA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[sjhalayka]

I found a happy medium: uintptr_t* genovec = (uintptr_t*)malloc(sizeof(uintptr_t) * 64); I'm absolutely stumped as to why it works, but the absolute minimum number of elements in the genovec array seems to be 64. I allocated more than that, filled it with the sentinel value -9999, and ran it through the PgrGet function. After that I printed out the genovec array and found that the overwritten size is always 64, regardless of the sample and variant counts and chunk_size. Why 64 though, I wonder? - Shawn

…

On Wed, Mar 5, 2025 at 10:04 AM Shawn Halayka ***@***.***> wrote: I tried that and it gives me a core dump because genovec is too small. Thanks for helping though! I sincerely appreciate it. On Tue, Mar 4, 2025 at 11:52 PM Christopher Chang < ***@***.***> wrote: > fyi, the genovecs array should have NypCtToWordCt(sample_ct) elements. > > — > Reply to this email directly, view it on GitHub > <#293 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AD5FASU3SSADW57HVSIP7UL2S2GJNAVCNFSM6AAAAABYJV4PNGVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTEMZZGY4DEMA> > . > You are receiving this because you authored the thread.Message ID: > ***@***.***> >

[chrchang]

I should have clarified that NypCtToWordCt(sample_ct) is the number of uintptr_ts. You need to multiply that number by sizeof(uintptr_t) to get the correct number of bytes.

sizeof(uintptr_t) * 64 will stop working when you start testing with larger sample_ct.

[sjhalayka]

Ok, I will revert to not using 64. Thank you for all of your guidance.On Mar 5, 2025, at 10:41 AM, Christopher Chang ***@***.***> wrote: I should have clarified that NypCtToWordCt(sample_ct) is the number of uintptr_ts. You need to multiply that number by sizeof(uintptr_t) to get the correct number of bytes. sizeof(uintptr_t) * 64 will stop working when you start testing with larger sample_ct. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>