You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Replace realloc(3) and maybe malloc(3) with std::vector #10813
... and the new operator! Just had a look at IdoMysqlConnection#Escape() and thought: srsly? new char[] and delete []? This could be a RAII std::vector!
Plan is to replace manual realloc/malloc buffer growth in IdoMysqlConnection::Escape() with std::vector<char> while keeping output and error handling unchanged. If there is a nearby test target for this path, I will add coverage in the same PR.
@qorexdevs You have already created the PR #10821 for this issue, didn't sign the CLA and closed the PR again. Please convince me that you're not a bot going around in circles, otherwise, please stop.
Fair point. PR #10821 was a separate attempt, and I closed it because the CLA was still blocking review. This issue is still open, and I claimed it so I can work the current code path in one branch once the CLA is sorted. If you'd rather I step back until the CLA is signed, I'll do that.
I was planning to replace the manual memory management with RAII containers in Process::IOThreadProc() and IdoMysqlConnection::Escape(), as mentioned in the comments. Is that the intended scope for this issue? If so, I'd be happy to work on it.
Activity
Al2Klimov commentedon Apr 24, 2026
... and the new operator! Just had a look at
IdoMysqlConnection#Escape()and thought: srsly?new char[]anddelete []? This could be a RAIIstd::vector!qorexdevs commentedon May 22, 2026
I can take this one.
Plan is to replace manual
realloc/mallocbuffer growth inIdoMysqlConnection::Escape()withstd::vector<char>while keeping output and error handling unchanged. If there is a nearby test target for this path, I will add coverage in the same PR.julianbrost commentedon May 22, 2026
@qorexdevs You have already created the PR #10821 for this issue, didn't sign the CLA and closed the PR again. Please convince me that you're not a bot going around in circles, otherwise, please stop.
qorexdevs commentedon May 22, 2026
Fair point. PR #10821 was a separate attempt, and I closed it because the CLA was still blocking review. This issue is still open, and I claimed it so I can work the current code path in one branch once the CLA is sorted. If you'd rather I step back until the CLA is signed, I'll do that.
aditya-prasad1 commentedon Jul 1, 2026
Hi @julianbrost @Al2Klimov! I'd like to work on this issue.
I was planning to replace the manual memory management with RAII containers in
Process::IOThreadProc()andIdoMysqlConnection::Escape(), as mentioned in the comments. Is that the intended scope for this issue? If so, I'd be happy to work on it.